Open Standard for the Regulation of Artificial Intelligence Creativity

Foreword

The rise of artificial intelligence (AI) in digital media creation is blurring the lines between human and machine-generated content, creating an environment where transparency about creative origin is paramount. While this evolution offers significant creative potential, the accompanying lack of transparency poses critical challenges; trust in digital media can be eroded, the perceived value of human artistry diminished, and an inequitable environment for creators and consumers fostered.

RAIC directly addresses these challenges. It provides a clear, consistent, and reliable framework for understanding the role of AI in digital works. By establishing a common language and a verifiable certification process, RAIC facilitates clear communication by creators about their work’s origins, encourages the ethical and responsible integration of AI in creative fields and empowers consumers to make informed choices.

Crucially, RAIC is a tool for transparency, not evaluation. It does not seek to judge the merit of a work based on its classification. Instead, it aims to cultivate a more informed and equitable digital landscape by acknowledging the spectrum of human-AI collaboration – from purely human artistry to predominantly AI-driven creation. RAIC promotes a future where AI serves to augment, not undermine, human creativity. This fosters responsible, ethical, and transparent practices.

Lumi Foundation is a non-profit organization that offers this standard as a contribution to the critical discussion surrounding AI’s societal impact. We believe collaborative efforts are essential to shaping a future where technology and human ingenuity synergistically create a richer and more vibrant digital world.

1. Introduction

RAIC is a framework designed to address the growing need for transparency in the creation of digital works involving artificial intelligence (AI). This standard provides a classification system that allows creators to communicate the extent of human and AI involvement in their work, enabling consumers to make informed decisions based on their preferences.

The RAIC standard is built upon the following core principles:

• Transparency: Providing clear and honest information about the role of AI in the creation of digital works.
• Empowering Consumer Choice: Giving consumers the information they need to make informed decisions based on their preferences regarding human vs. AI creativity.
• Embracing Subjectivity: Recognizing that evaluating creative contribution inherently involves subjective judgment, and that this subjectivity is essential for capturing the nuances of human-AI collaboration.
• Focusing on Process, Not Quality: RAIC is not about judging the artistic merit or quality of a work, but rather about describing the process by which it was created. A work primarily made by AI is not inherently “better” or “worse” than a human-made one - they are simply different.
• Adaptability: The standard is designed to be adaptable to the rapidly evolving landscape of AI and its use in creative fields.
• Fairness and Accountability: The certification process is designed to be fair and transparent, with mechanisms in place to address disputes and ensure accountability.

1.1 Scope

This standard is voluntary and can be used by anyone. It focuses on classifying complete digital works intended for public consumption where the primary intent is creative or artistic expression. This includes works intended to be experienced as a cohesive whole, rather than individual components or tools used in their creation.

1.1.1 Included Works

Examples of included works:

• Video games
• Digital animations
• Music (compositions and recordings)
• Films and videos (including short films and documentaries)
• Literary works (novels, short stories, poetry)
• Interactive media (VR/AR experiences)
• Digital paintings and illustrations
• Generative art installations
• Other forms of digital creative expression

1.1.2 Excluded Works

This standard specifically excludes works that are not considered to be complete creative products intended for public consumption. This includes:

• Technical and Operational Systems: Systems lacking primary creative intent, such as business software, operating systems, databases, industrial control systems, or other similar systems.
• Raw Data and Training Resources: Datasets, codebases, and training materials used in creation of a work, such as a database of images, sounds, or text used for training an AI model, a collection of motion capture data, a set of rules for a game, or a library of 3D models.
• Tools and Libraries: Software tools, libraries, and code enabling media creation, such as a digital audio workstation (DAW), a video editing software, a 3D modeling tool, or a game engine.
• Individual Digital Assets: Standalone assets are excluded unless incorporated into a larger work as a minor, non-substantial part. For instance, a single 3D model, a stock sound effect, or a texture, would be excluded if sold on their own. However, if these assets are part of a larger creative work, they can be evaluated as part of that work.
• Intermediate Work Products: Preliminary or in-process materials that are part of the production pipeline, but not part of the final creative work, are excluded. Examples include project files, raw footage, unmixed audio stems, isolated animation sequences, test renders, early drafts of a script or story, concept art or storyboards.
• Non-Public Works: Works created for private or personal use, not intended for public distribution are excluded.
• Educational and Research Projects: Works primarily for academic or AI research purposes are excluded, provided that they are not used for commercial release.

Note: While the above categories are excluded as standalone products, they can be part of a certified work. Such a certified work’s classification depends on the holistic use of contained works regardless of each work’s individual RAIC inclusion status.

1.2 Relationship to Other Standards

RAIC is designed to be complementary to, not a replacement for, standards focused on digital content attribution and provenance, such as the Content Authenticity Initiative (CAI) and C2PA, and will also be compatible with future standards in this space. While CAI and C2PA focus on verifying the authenticity and origin of individual digital assets (e.g., images, audio clips), RAIC addresses the broader creative process of a complete work.

RAIC can leverage the foundations established by these standards. For example, C2PA-compliant assets can serve as supporting evidence for RAIC official certification claims. RAIC focuses on the integration of various assets and tools – whether human-created or AI-generated – into a final, cohesive work. The standard emphasizes the holistic creative process rather than the provenance of individual components.

RAIC is designed to work alongside other standards that address different aspects of digital asset creation and usage. One such standard is the LACS Asset Classification Standard (LACS), which focuses on classifying individual digital assets based on the level of human and AI involvement in their creation (LACS-H for Human-Made, LACS-A for AI-Assisted, and LACS-G for AI-Generated).

While RAIC addresses the overall creative process of a complete work, LACS provides information about the origin of individual assets used within that work. LACS classification can serve as valuable supporting evidence for RAIC certification, particularly in demonstrating the use of AI-generated assets and their integration into the larger creative process. For example, a work containing LACS-G assets would likely not qualify for RAIC Class A certification.

Creators are encouraged to use LACS to classify their individual assets, as this can streamline the RAIC certification process and provide greater transparency to consumers. When submitting a work for RAIC certification, creators can provide information about the LACS classification of their assets, which will be considered by certifiers during the evaluation process.

1.3 Terms and Definitions

This section provides definitions for key terms used throughout the RAIC standard.

• AI-Assisted Creation: The use of AI-powered tools to aid a human creator’s creative process. This can involve generating variations, exploring options within human-defined parameters, or performing tasks where the human may lack expertise. The human maintains control over all key creative decisions and the AI does not independently shape the work’s artistic expression.
• AI-Generated Content: Content produced by an AI system, which can range from individual elements to entire works. The degree of human involvement in shaping AI-generated content varies depending on the context and the specific classification of the work.
• Algorithmic Element: An asset, technique, method, resource, or any other component that forms part of a work and that was created by algorithms.
• Certification: The process of assigning an RAIC label to a work.
• Creator: An individual or group responsible for creating a digital work, including game developers, musicians, filmmakers, writers, and artists.
• Creativity: The ability to come up with new ideas, concepts, and expressions that are original and meaningful. In the context of this standard, creativity involves making deliberate choices that shape the form, content, and meaning of a work.
• Creative Input: Contribution to the creative process that shapes the work’s artistic expression, content, meaning, or overall impact on the audience. This can involve making Creative Decisions, generating ideas, providing guidance and feedback, or any other action that influences the final form and essence of the work.
• Creative Decision: A choice that significantly impacts the work’s artistic expression, content, structure, user experience, or overall aesthetic and thematic impact. This can include decisions about style, composition, narrative elements, gameplay mechanics, selection of specific elements generated by an AI, and other aspects that shape the final form, meaning, and reception of the work.
• Algorithmic Element: An asset, technique, method, resource, or any other component that forms part of a work. Algorithmic elements are categorized as either inconsequential or consequential, depending on their creative significance.
• Artificial Intelligence (AI): AI systems are computer systems capable of generating or modifying content, or of making Creative Decisions. For the purposes of this standard, AI is considered in a broad sense, including but not limited to machine learning models, deep learning systems, and other forms of computational intelligence that can contribute to the creative process.
• Minimal Human Input: In the context of Class Z, minimal human input refers to actions such as providing initial prompts, setting parameters for AI systems, and curating AI-generated outputs. It does not involve modifying the AI-generated content or directly shaping the work’s artistic expression.
• Creative Intent: The artistic, narrative, functional, or experiential goals guiding a work’s creation, from concept to implementation, encompassing the original idea and all related aesthetic and technical design aspects.
• Human Oversight: Active human involvement in guiding and reviewing the creative process, particularly when using AI tools.
• Inconsequential AI Tool Use: Use of AI for tasks that are unrelated to the creative process, such as software bug fixes, code formatting, or data organization. This is always allowed, regardless of the work’s classification.
• Licensed Certifier: A third-party organization approved by the Lumi Foundation to conduct official RAIC certifications.
• Official Certification: The process of verifying a work’s adherence to a specific RAIC classification through an independent evaluation by a Licensed Certifier, resulting in an official RAIC certification. The Licensed Certifier will primarily base their assessment on interviews, documentation, and demonstrations provided by the creator. While access to source code or proprietary assets is generally not required, the certifier may request specific, targeted evidence if needed to verify claims. However, this will be limited to the minimum necessary to resolve any doubts or discrepancies, and creators are not obligated to disclose their complete source code nor any proprietary assets.
• RAIC ID: A unique alphanumeric code assigned by the Lumi Foundation to each certified work upon completion of the certification process. The RAIC ID is composed of seven alphanumeric characters, excluding 0 (zero), 1 (one), O, I, because of visual similarities. This identifier is linked to the work’s detailed certification information, including its classification, within the RAIC database. (See Section 4.4 for more information about the structure of the RAIC ID).
• Self-Certification: The process where the creator of a work receives an RAIC classification based on their answers to an online guided questionnaire. This process is less rigorous than official certification but still requires creators to act in good faith and adhere to the RAIC guidelines. Self-certification results in the work being added to the RAIC database.
• Source Asset: A core element of the final work (e.g., source asset, images, 3D models, sound effects, musical phrases, text blocks, code) that is a key component of the finished product.
• Work: In the context of this standard, “work” refers to a complete and cohesive creative product intended for distribution and consumption, such as a video game, musical piece, or film.
• Label or RAIC Label: A designation that represents the classification of a work (e.g., “Class A”, “Class B”).
• Badge or RAIC badge: The visual logo that can be displayed on products that have received an RAIC certification. There is a unique badge for each class.
• Creative Process Documentation: A comprehensive document that details both the human contributions and the use of AI in the creation of a work. It should list all individuals involved in the creative process and their respective roles, as well as specify which parts of the project were created or assisted by AI, the type of AI tools used, and the extent of AI involvement. This document is crucial for determining the RAIC classification of a work.

1.4. Algorithmic Elements Definitions

The concept of algorithmic elements are used throughout this document to denote assets, techniques, methods, resources, and more, that form parts of a Work. For the purposes of this standard, elements are categorized as inconsequential or consequential, depending on their creative significance.

• REQ-ELEM-001: Inconsequential Algorithmic Elements (IAEs) are algorithmic elements that do not significantly contribute to the work’s unique creative expression, aesthetic, narrative, or experiential qualities. They are often elements that require minimal creative input, are easily replaceable without impacting the overall artistic intent, or serve a purely functional purpose. Their method of creation, whether through algorithms, automated processes, or by an AI, is irrelevant to the work’s classification, as long as they remain inconsequential to the overall Creative Intent. Using AI tools to create IAEs is allowed in all RAIC classes and does not affect the classification.
• REQ-ELEM-002: Consequential Algorithmic Elements (CAEs) contributes meaningfully to the creative intention of a work. Unlike IAEs, they are not merely basic elements requiring minimal creative input, but play a meaningful role in shaping the work’s overall form, content, or meaning.

To aid in determining whether an element is consequential, creators and certifiers must consult the RAIC Precedent Database. This database provides examples of how the CAE/IAE distinction, which inherently involves a degree of subjectivity (see Section 1.5), has been applied in various contexts.

1.5. Subjectivity

The RAIC standard acknowledges that the determination of whether an algorithmic element is consequential (CAE) or inconsequential (IAE) involves a degree of subjectivity. This is not a flaw, but rather a reflection of the inherent nature of evaluating creative contributions. Unlike purely technical standards, RAIC deals with the nuanced interplay between human and artificial intelligence in the creative process, a domain where artistic significance and creative intent are paramount.

Why Subjectivity is Necessary

• Artistic Judgment: Assessing the creative impact of an element often requires artistic judgment, which is inherently subjective. What one person considers a significant contribution, another might view as a minor detail. This variability is a natural part of the creative world.
• Context-Dependence: The significance of an element can vary greatly depending on the specific context of the work. A seemingly simple pattern might be inconsequential in one work but a central motif in another.
• Evolving Nature of AI: As AI tools become more sophisticated, the line between human and AI contributions may become increasingly blurred. A degree of subjectivity allows the standard to adapt to these advancements.
• Empowering Consumers: Ultimately, RAIC aims to empower consumers to make informed choices based on their own values and preferences regarding human vs. AI creativity. Subjectivity in the standard mirrors the subjectivity in consumer preferences, allowing for a more nuanced and meaningful evaluation of creative works.

Embracing Subjectivity

Rather than striving for an unattainable objectivity, RAIC embraces subjectivity as a feature that enhances its relevance and aligns it with the realities of the creative landscape. The standard provides mechanisms to manage this subjectivity responsibly:

• Guided Questionnaire: The self-certification process utilizes a guided questionnaire designed to elicit nuanced information about the creative process, allowing for a more accurate classification despite the subjective nature of the assessment.
• Licensed Certifiers: Official certification relies on the expertise of licensed certifiers, who are trained to evaluate creative works and make informed judgments about the significance of algorithmic elements.
• Certification Review Board: The Certification Review Board serves as the final arbiter in cases of disputes or appeals, providing a mechanism for addressing borderline cases and establishing precedents.
• Transparency and Documentation: RAIC emphasizes transparency throughout the certification process. Certifiers are required to document their reasoning, and the Certification Review Board’s decisions are made public, creating a body of knowledge that guides future evaluations.
• Precedent Database: The Precedent Database serves as a crucial resource for managing subjectivity. By providing concrete examples of how the CAE/IAE distinction has been applied in various contexts, the database helps to establish a shared understanding of what constitutes a significant creative contribution. While not a rigid set of rules, the database provides a framework for making informed and consistent judgments.

By acknowledging and managing subjectivity, RAIC provides a framework for understanding the complex interplay between human and AI in the creative process. A key component of this framework is the Precedent Database, a searchable repository of decisions made by the Certification Review Board on borderline cases. This database serves as a crucial resource for both creators and certifiers, providing concrete examples and rationales to guide their assessments of algorithmic elements. The Precedent Database, combined with the expertise of licensed certifiers and the guided questionnaire, allows RAIC to be adaptable and relevant in the evolving landscape of digital art, ultimately empowering both creators and consumers.

• REQ-SUBJ-007: The determination of whether an algorithmic element is consequential (CAE) or inconsequential (IAE) is ultimately based on the collective judgment of the community, as reflected in the precedents established by the Certification Review Board and documented in the Precedent Database. This collective judgment evolves over time as new cases are considered and new precedents are set.

2. Classification Framework

2.1. Overview of Classes

The RAIC classification system categorizes digital works based on the degree of human and AI involvement in their creation. It uses a letter-based system (A, B, U, Y, Z, and F) to represent different levels of AI involvement. These classifications are descriptive, not prescriptive, and aim to provide information about the creative process without making value judgments about the quality or merit of the work.

2.1.1 Legal and Ethical Compliance

• REQ-RAIC-001: All works, regardless of class, must respect existing copyright and intellectual property laws. This includes obtaining necessary licenses for any third-party content used, and ensuring that the work does not infringe on the rights of others.

2.2. Class A: Human Craftsmanship

Description: Class A represents works where human Creativity is the sole driving force. Human artists make all significant creative decisions, and the work is free of any consequential use of AI and all CAEs, regardless of their origin. This category highlights purely human-driven creative processes.

2.2.1 Content Creation Requirements

• REQ-A-000: Class A works are allowed to incorporate IAEs, as defined in REQ-ELEM-001, even if those elements are created using AI tools.
• REQ-A-001: All source assets must be verifiably created by human artists.
• REQ-A-002: Source assets in Class A works cannot be based on AI-generated CAEs or modifications thereof.
• REQ-A-003: The creation of visual elements in the final work must be the direct result of human artistry through methods such as drawing, painting, sculpting, modeling, or other manual methods.
• REQ-A-004: Audio elements must be directly created by human artists through recording, composition, performance, or manual production.
• REQ-A-005: Text-based content and narrative elements must be directly created by human authors.
• REQ-A-006: Software and code that form a CAE part of the work must be directly programmed or designed by a human, without using AI to generate code.

2.2.2 Asset Requirements

• REQ-A-021: Proof of non-AI origin can come from:
  • Certification by a trusted standards body,
  • A verifiable declaration from the original asset creator,
  • Documentation of the creative process, or
  • Other reasonable evidence of human creation.
• REQ-A-023: Asset integration must be a result of direct human creative decisions.

2.3. Class B: Human-Directed AI Enhancement

Description: Class B represents works where human creators maintain primary creative control over the work’s Creative Intent and direction. AI may be used to generate CAEs that are included in the final work, but these AI contributions must be in direct response to specific human creative direction and serve to enhance or augment the human’s Creative Intent. The key distinction from Class A is the explicit allowance of AI-generated CAEs, under specific conditions of clear, documented, and demonstrably significant human direction.

2.3.1. Content Creation Requirements

• REQ-B-001: The core Creative Intent, artistic direction, and overall aesthetic of the work must originate from and be driven by human creators.
• REQ-B-002: AI-generated CAEs are permitted when they are used to implement specific, well-defined human creative goals. The human creator must provide detailed direction and specifications to guide the AI’s output.
• REQ-B-003: AI-generated CAEs must be integrated into the work in a way that supports and enhances the human’s Creative Intent. The final work should reflect a unified artistic style and intent that is primarily shaped by the human creator.
• REQ-B-004: The use of AI tools must be documented, demonstrating the specific prompts, parameters, and settings used to generate CAEs. This documentation should clearly show the human’s role in directing the AI and shaping its output.
• REQ-B-005: Humans must make all final decisions about the selection, refinement, and integration of AI-generated CAEs, based on their artistic judgment and the overall Creative Intent for the work.
• REQ-B-006: While Artificial Intelligence may be used to generate variations or options, the final choice and implementation of elements within the work must be made by the human creator, demonstrating a conscious and deliberate creative decision-making process.

2.3.2. AI Usage Requirements

• REQ-B-010: AI tools may be used to generate CAEs under specific human direction.
• REQ-B-011: AI generation of CAEs must be guided by detailed creative specifications that determine the parameters of the AI’s output, instead of non-specific prompts or exploratory methods.
• REQ-B-012: Humans must make all final decisions about the selection and refinement of CAEs, based on their artistic judgment.
• REQ-B-013: The integration of CAEs must serve the overall Creative Intent that is created by the human artists.

2.3.3. Human-AI Interaction

• REQ-B-020: Humans must maintain creative control over all AI-generated content, and how it is used in the work.
• REQ-B-021: The quality, suitability, and artistic merit of all parts of the project must be defined and evaluated by human creators.

2.4. Class U: Unified Human-AI Collaboration

Description: Class U represents works created through a collaborative partnership between human and AI, where both contribute significantly. The human makes the key creative decisions, even if relying on AI assistance due to skill gaps or for exploring creative options within human-defined parameters. The AI can have some autonomy within human-defined parameters but does not make independent creative decisions that significantly shape the work’s artistic expression without human guidance.

2.4.1. Basic Requirements

• REQ-U-001: Both human and AI contribute significantly to the work. The human makes the key creative decisions, while the AI can provide assistance and explore options within human-defined parameters.
• REQ-U-002: The final work must reflect the human’s Creative Intent, even if AI tools are used extensively for assistance or exploration.

2.4.2. Creative Process

• REQ-U-010: Creative decisions originate from the human, who defines the Creative Intent and goals.
• REQ-U-011: AI systems can be used to assist the human by generating variations, exploring options, or performing tasks where the human may lack expertise. The AI’s suggestions are always subject to human approval and must align with the human’s overall Creative Intent.

2.5. Class Y: AI-Driven with Human Oversight

Description: Class Y represents works where AI systems take a leading role in creative decisions, with humans providing high-level guidance and oversight. This classification reflects a workflow where the AI is responsible for generating significant portions of the work’s content and making many of the creative choices, while humans set the overall direction, define goals, and ensure the AI’s output aligns with their Creative Intent.

2.5.1. Basic Requirements

• REQ-Y-001: AI systems are responsible for making many of the creative decisions, shaping the work’s content and artistic expression.
• REQ-Y-002: Human involvement includes setting high-level creative goals, defining project-specific guidelines, and providing feedback to steer the AI’s creative process. Humans may also curate or refine AI outputs but do not necessarily make detailed modifications.
• REQ-Y-003: Humans are responsible for ensuring the AI’s output meets the desired quality standards and aligns with the overall Creative Intent for the work.

2.5.2. Human Role

• REQ-Y-010: Humans define the overall Creative Intent, set high-level goals, and provide project-specific guidelines for the AI.
• REQ-Y-011: Humans provide feedback to guide the AI’s creative process and may curate or refine AI outputs to ensure they align with the overall Creative Intent.

2.6. Class Z: Artificial Craftsmanship

Description: Class Z represents works that are primarily generated by AI systems, with minimal human input. In this classification, AI systems are responsible for both the conceptualization and realization of the work, making the vast majority of creative decisions. Human involvement is limited to providing initial prompts, setting parameters, and curating the AI-generated outputs.

2.6.1. Basic Requirements

• REQ-Z-001: AI systems are the primary creative force, responsible for generating the work’s content and making the vast majority of creative decisions.
• REQ-Z-002: Human involvement is limited to providing initial prompts, setting parameters, and curating the AI-generated outputs. Humans do not modify the AI-generated content or directly shape the work’s artistic expression.

2.6.2. Legal and Ethical Compliance

• REQ-Z-003: The work must still meet basic quality and coherence standards.

2.7. Class F: Fraudulent or Misleading

Description: Class F is a special designation reserved for works that are found to have significantly misrepresented their creative process, or otherwise engaged in deceptive practices related to RAIC certification. This classification is not assigned during the initial certification process, but rather as a result of an investigation triggered by community reports, audits, or other forms of scrutiny. It is designed to protect the integrity of the RAIC standard and to deter deliberate misrepresentation. The Class F designation will only be applied after a thorough investigation and appeals process, as outlined in Section 5.4 and 6.3.

2.7.1. Conditions for Designation

• REQ-F-001: A Class F designation is applied when an investigation concludes that a work has been fraudulently certified or has significantly misrepresented the role of AI or human involvement in its creation.
• REQ-F-002: Misrepresentation can include, but is not limited to:
  • Stating a higher level of human involvement than is demonstrable.
  • Using CAEs while claiming Class A status.
  • Providing false or misleading documentation, or otherwise misrepresenting the creative process (only applicable in official certifications).
  • Failing to disclose the use of AI tools when required by the claimed classification.
  • Using the RAIC certification without any form of documentation or proof of origin.
  • Using the RAIC certification in an illegal manner, or otherwise abusing the intended use of the certification.
• REQ-F-003: Class F designation can be applied to any work, regardless of its initial self-certified or officially certified status. It overrides any previous classification.
• REQ-F-004: The decision to assign a Class F designation is made by the Lumi Foundation, based on the recommendation of the Certification Review Board after a thorough investigation and appeals process (see Sections 5.4 and 6.3).

2.7.2. Documentation and Transparency

• REQ-F-010: A detailed report justifying the Class F designation must be created and made publicly available. This report must include:
  • Specific evidence of misrepresentation or fraud.
  • A clear explanation of how the work violates RAIC requirements.
  • A record of the investigation process, including communication with the creators or distributors.
• REQ-F-011: The documentation must be sufficient to allow independent verification of the Class F designation.
• REQ-F-012: The report must be published in the RAIC certification database, linked to the work’s unique identifier, and to any formal statements made about the project.

2.7.3. Right to Defense and Appeal

• REQ-F-020: Before any work is assigned a Class F designation, the creator or rights holder will be given a chance to respond to the allegations and provide their own evidence. They will be provided with a detailed report outlining the alleged misrepresentation and the evidence gathered against them.
• REQ-F-021: The Lumi Foundation will establish a clear process for creators to present their case, including timelines and procedures for submitting evidence. This process will be overseen by the Certification Review Board and will be designed to be fair, impartial, and transparent.
• REQ-F-022: Only after this process is complete, and the evidence has been reviewed, will a final decision on Class F designation be made by the Certification Review Board. The Certification Review Board will provide a detailed rationale for their decision, addressing the creator’s arguments and evidence.

2.8. Managing Subjectivity in Certification

• REQ-SUBJ-001: As discussed in Section 1.5, judging whether an element is a CAE or an IAE involves a degree of subjectivity. This section outlines how subjectivity is managed within the certification process.
• REQ-SUBJ-002: In self-certification, the creator answers the guided questionnaire honestly, and the system classifies based on their responses. This relies on the creator’s good faith and understanding of the CAE/IAE distinction as explained in Section 1.5. Creators are encouraged to consult the Precedent Database for guidance.
• REQ-SUBJ-003: In official certification, licensed certifiers make independent assessments, informed by their expertise and the principles outlined in Section 1.5. While certifiers strive for objectivity, their judgments may still involve subjective elements. They are accountable for their decisions and must adhere to the RAIC standard.
• REQ-SUBJ-004: The Certification Review Board serves as the final arbiter in disputes or appeals, providing a mechanism for managing subjectivity at the highest level. Their decisions are based on evidence, the RAIC standard, and their collective judgment, as detailed in Section 1.5.
• REQ-SUBJ-005: All parties are expected to act in good faith, aiming for fair and accurate assessments.
• REQ-SUBJ-006: The determination of CAE or IAE ultimately rests with the creator (self-certification), the licensed certifier (official certification), or the Certification Review Board (disputes). While guidelines and precedents are provided, the final decision requires careful consideration of the specific context, as emphasized in Section 1.5.

3. Certification Process

3.1. General Requirements

3.1.1. Basic Principles

• REQ-CERT-001: All certification processes must prioritize transparency and verifiability.
• REQ-CERT-003: Certification must cover the entire work, including all significant assets and components.
• REQ-CERT-004: The official certification process must be repeatable and based on verifiable evidence.
• REQ-CERT-005: In self-certification, creators bear sole responsibility for the accuracy of their submissions. During official certification, this responsibility shifts to the certifier for their evaluations.
• REQ-CERT-006: All documentation must be clear, well-organized, and sufficient to allow the certifier to verify the accuracy of the classification claims.
• REQ-CERT-007: The documentation should be submitted in a commonly accessible format (e.g., PDF, plain text, common image formats).

3.1.2. Documentation Requirements for Official Certification

• REQ-DOC-010: Official certification mandates a Creative Process Documentation detailing both human and AI contributions. This document should include a list of all contributors and their roles, and a thorough description of AI involvement, specifying tools used and their impact on the creative process. The goal is to enable certifiers to differentiate between CAEs and IAEs.
• REQ-DOC-011: The information provided in the Creative Process Documentation should be high-level enough to provide a general overview of the creative process but detailed enough to allow the certifier to make an informed decision about the work’s classification. Creators aiming for either Class A or Class Z certification should be particularly thorough in their documentation to demonstrate either the complete absence of AI-generated CAEs (Class A) or the predominant use of AI in the creative process (Class Z).
• REQ-DOC-012: While the Creative Process Documentation is usually sufficient, certifiers may request further details, such as AI output examples or LACS classifications of assets, to ensure accurate assessment.
• REQ-DOC-013: In exceptional cases, such as when there is a dispute about the origin of a particular element or when the certifier suspects misrepresentation, the certifier may request access to specific portions of source material or assets. This will only be done when absolutely necessary to verify claims and after providing a clear justification to the creator. The certifier must handle any such materials with strict confidentiality and in accordance with legal requirements. The Creator may require the certifier to sign a Non-Disclosure Agreement.

Note: Determining whether an element is a CAE or an IAE involves a degree of subjectivity, as discussed in Section 1.5. The certifier will make the assessment based on the provided documentation, and the Certification Review Board makes the final determination in cases of disputes or appeals.

3.2. Self-Certification Process

3.2.1. Overview

• REQ-SELF-001: Self-certification is an automated process designed to guide creators through a series of questions that evaluate their work’s creative process and determine its RAIC classification.
• REQ-SELF-002: Creators do not claim a class but rather receive one based on their honest answers to the guided questionnaire.

3.2.2. Guided Questionnaire

• REQ-SELF-010: The self-certification process consists of an online, interactive questionnaire that uses a decision-tree logic.
• REQ-SELF-011: The questionnaire will cover human involvement in ideation, design, and execution; use of AI tools and their roles; the nature of CAEs and IAEs; and the overall creative control process.
• REQ-SELF-012: Based on the creator’s responses, the questionnaire will automatically determine the appropriate RAIC classification (A, B, U, Y, or Z).
• REQ-SELF-013: The questionnaire will be meticulously designed to minimize ambiguity and ensure accurate classification based on the RAIC standard’s criteria.

3.2.3. Importance of Understanding CAE/IAE Distinction in Self-Certification

Before beginning the self-certification process, creators must understand distinction between CAEs and IAEs as defined in Section 1.4. This understanding is crucial for accurately answering the guided questionnaire.

The RAIC Precedent Database, a publicly available, searchable repository of decisions made by the Certification Review Board on borderline cases, can be a helpful resource for understanding the CAE/IAE distinction. While not mandatory for self-certification, creators are encouraged to consult the database, particularly if they are unsure about the classification of specific elements in their work. The database contains detailed information on various algorithmic elements that have been classified as either CAEs or IAEs in specific contexts, along with the rationale behind these decisions.

The Precedent Database can be accessed through the RAIC website and offers advanced search functionality, including keyword search and filtering by element type. While self-certification relies on the creator’s good faith and understanding of the CAE/IAE distinction, the Precedent Database can provide valuable insights and clarification, especially for complex or borderline cases.

3.2.4. Certification Issuance

• REQ-SELF-020: Upon completion of the questionnaire, the RAIC classification will be automatically assigned and added to the RAIC database along with basic information about the work (title, creator, type of work, etc.).
• REQ-SELF-021: Creators will receive a unique RAIC ID for their work, which they can use to display the appropriate RAIC badge.

3.2.5. Terms of Use

• REQ-SELF-030: Before starting the self-certification process, creators must agree to the terms of use, which include:
  • A declaration that they will answer all questions truthfully and to the best of their ability.
  • An acknowledgment that providing false or misleading information may result in penalties, including a Class F designation and blacklisting.
  • An understanding that they may face legal action from third parties, including the Lumi Foundation, for trademark misuse if they misrepresent their work.
  • Confirmation that they have the legal right to use the RAIC certification badges and any included assets and components.
  • An understanding that the accuracy of the self-certification relies on their honesty and that official certification is available for a more rigorous evaluation.

3.3. Official Certification Process

flowchart LR
    A[Start] --> B(Submit Application to Licensed Certifier);
    B --> C{Is it a borderline case?};
    B --> D(Certifier Reviews Documentation, Conducts Interviews);
    C -- Yes --> E(Refer to Certification Review Board);
    E --> F[Board Makes Final Determination];
    F --> D;
    C -- No --> D;
    D -- Assesses CAE/IAE --> G{Certifier Makes Classification Decision};
    G -- Class A, B, U, Y or Z --> H(Recommendation to Lumi Foundation);
    H --> I[Lumi Foundation Approves/Rejects];
    I -- Approves --> J[Official RAIC ID Issued & Report Published];
    I -- Rejects --> K[Rejection with Explanation];
    style A fill:#ccf,stroke:#333,stroke-width:2px
    style B fill:#fff,stroke:#333,stroke-width:2px
    style C fill:#ffc,stroke:#333,stroke-width:4px
    style D fill:#fff,stroke:#333,stroke-width:2px
    style E fill:#fcf,stroke:#333,stroke-width:2px
    style F fill:#fcf,stroke:#333,stroke-width:2px
    style G fill:#ffc,stroke:#333,stroke-width:4px
    style H fill:#fff,stroke:#333,stroke-width:2px
    style I fill:#ccf,stroke:#333,stroke-width:2px
    style J fill:#cfc,stroke:#333,stroke-width:2px
    style K fill:#fcc,stroke:#333,stroke-width:2px
    classDef highlight fill:#ffc,stroke:#333,stroke-width:4px

3.3.1. Introduction

Official certification provides a higher level of assurance than self-certification. It involves an independent evaluation of the work and its creative process by a Lumi Foundation licensed certifier. While it is a more rigorous and costly process, it provides a stronger validation of the work’s classification and can enhance consumer trust. Official certification also allows the creator to use a special version of the RAIC logo with an added asterisk, signifying the official certification.

3.3.2. Eligibility

• REQ-CERT-101: The work must be in final or near-final form, and no major modifications must occur during the certification process.
• REQ-CERT-102: Sufficient materials for evaluation must be available for the certification process.

3.3.3. Licensed Certifiers

• REQ-CERT-110: Official certification must be performed by licensed RAIC certifiers that are approved by the Lumi Foundation.
• REQ-CERT-111: To become a licensed certifier, an organization must demonstrate:
  • Precedent Proficiency: Demonstrated ability to effectively utilize the RAIC Precedent Database when making classification decisions, including understanding how to search for relevant precedents and apply them to new cases.
  • Expertise: Relevant technical expertise in AI, software development, and creative processes relevant to the types of works they intend to certify.
  • Impartiality: A commitment to objectivity and impartiality, with no conflicts of interest that could compromise their judgment.
  • Ethical Conduct: A commitment to ethical conduct and adherence to the principles of the RAIC standard.
  • Experience: Demonstrable experience in auditing, verification, or a related field.
  • Capacity: Sufficient resources and personnel to conduct certifications in a timely and efficient manner.
  • Reputation: A demonstrable track record of ethical conduct and reliability in their respective field.
  • Legal Responsibility: A clear understanding and acceptance of their legal responsibility for the accuracy and fairness of their assessments. They must also acknowledge their liability for any damages caused by negligence or misconduct in the certification process.
• REQ-CERT-112: All certifiers must maintain full independence from the works they evaluate to guarantee neutrality and objectivity.
• REQ-CERT-113: The Lumi Foundation will establish a formal licensing process for certifiers, which will include an application, vetting, training, and ongoing quality control.
• REQ-CERT-114: Licensed certifiers will be listed on the Lumi Foundation website, together with their unique 4-digit identifier.
• REQ-CERT-115: Licensed certifiers are authorized to use the official RAIC certification badges in connection with their services, subject to the guidelines outlined in Appendix C. They will also be awarded a special Licensed RAIC Certifier icon, which they can use on their website or marketing material.
• REQ-CERT-116: Certifiers are required to maintain detailed records of their evaluations, including all communications with creators, documentation reviewed, and the rationale for their certification decisions.
• REQ-CERT-117: Certifiers are legally responsible for their evaluations and are liable for any damages caused to creators due to negligence, misconduct, or breaches of confidentiality. The Lumi Foundation will actively support creators in pursuing legal action against certifiers who violate these terms. This includes providing legal assistance and resources to help creators protect their rights and interests.
• REQ-CERT-118: Certifiers are subject to periodic audits by the Lumi Foundation to ensure the quality and consistency of their evaluations. As part of this process, individual works that have been certified by a certifier will be randomly selected for review by a different, randomly assigned certifier. If the reviewing certifier disagrees with the original certification decision, they must submit a report to the Certification Review Board outlining their concerns. The Certification Review Board will then make a final determination on the validity of the original certification.

3.3.4. Application Process

• REQ-CERT-120: An application for official certification can be initiated either by submitting a request to the Lumi Foundation or by directly contacting a licensed certifier. Regardless of the initiation method, the application process requires the following information:
  • Basic information about the work (title, creator, type of work, etc.).
  • The RAIC ID, if available.
  • Documentation according to 3.1.2.
• REQ-CERT-121: The evaluation has a fixed fee of USD $99 (excluding VAT), and the certifier is expected to perform all tasks required to make a determination, including all communication with the applicant, all necessary analysis, and any needed research. The certifier must also disclose any potential conflicts of interest.

3.3.5. Evaluation Process

• REQ-CERT-130: The certifier will conduct interviews with the creator and key personnel involved in the production process, and they will also send project-specific questions, based on the description, and a thorough review of the documentation. The certifier may also, in exceptional cases, request access to specific portions of source materials or assets if deemed necessary to verify claims, but only after providing a clear justification. The goal is to verify all claims, and to determine whether elements are CAEs or IAEs.
• REQ-CERT-130: In making these determinations, the certifier will consult the RAIC Precedent Database, searching for similar cases and applying the principles established by the Certification Review Board. The certifier will document their use of the database, noting any relevant precedents that informed their decision.
• REQ-CERT-131: The certifier might also request additional documentation or clarifications when needed to understand key creative or technical choices.
• REQ-CERT-132: The certifier evaluates the production pipeline, tools, assets, and methods to classify the project, determining the nature of elements as CAEs or IAEs. Borderline or complex cases are referred to the Certification Review Board. The evaluation must be thorough, within the fixed fee, irrespective of project complexity.
• REQ-CERT-133: The evaluation process will specifically focus on:
  • Assessing the extent of human involvement and the degree to which creative decisions are made by humans versus AI systems.
  • Determining the degree of AI autonomy in the creative process, particularly in distinguishing between Class U, Y, and Z.
• REQ-CERT-134: The certifier will primarily rely on the submitted Creative Process Documentation to conduct their evaluation. The certifier will use this information to determine the appropriate RAIC classification (A, B, U, Y, or Z) for the work, based on the extent of human and AI involvement in the creative process. The assigned classification is based on the certifier’s assessment and is not chosen or claimed by the creator. Further review may involve:
  • Interviews: Discussions with the creators to understand their creative process, decision-making, and the specific roles of humans and AI.
  • Demonstrations: Creators may be asked to demonstrate how they used AI tools in the creation process, showcasing specific examples of human input and AI output.
  • Targeted Evidence Requests: Certifiers may request specific pieces of evidence to support claims, such as design documents, early drafts, or specific code snippets, but always within reason.
  • Documentation Review: Examining any additional documentation provided by the creator, such as design documents, process descriptions, or AI tool usage logs.

Note: The evaluation process aims to respect the intellectual property and confidentiality concerns of creators. The certifier will not require access to the complete source code or all proprietary assets of a work unless absolutely necessary for verification. The evaluation will primarily rely on the Creative Process Documentation, supplemented by interviews, targeted evidence requests, and further documentation review if needed. The certifier will be required to present a strong justification for requesting access to any part of the work, and will have to demonstrate why such access is absolutely necessary to make a fair evaluation. Certifiers are also required to clearly state how such materials will be used, handled, stored, and eventually disposed of. Certifiers are legally liable for any damages caused by their failure to adhere to these principles, and the Lumi Foundation will fully support creators in any such legal proceedings to ensure that creators are protected from malicious actions.

Note: As explained in Section 1.5, the determination of whether an element is a CAE or an IAE involves a degree of subjectivity. The certifier will make the determination based on the provided documentation, interviews, and their assessment of the work. In borderline cases or cases of disagreement, the certifier must refer the case to the Certification Review Board (see Section 5.4).

3.3.6. Certification Decision and Approval

• REQ-CERT-140: Certifiers recommend approval or rejection of certification to the Lumi Foundation. Recommendations must include the Certification Review Board’s decisions on borderline cases and a rationale for their impact on classification.
• REQ-CERT-141: The recommendation includes a detailed report with the proposed RAIC classification (A, B, U, Y, or Z), describing identified issues, including legal ones. It must address the reasoning for classifying borderline elements as CAEs or IAEs, applying RAIC standard definitions and criteria.
• REQ-CERT-142: The Lumi Foundation will review the certifier’s recommendation and make the final decision on whether to grant the RAIC certification.
• REQ-CERT-143: The Lumi Foundation will typically approve the certification if the certifier’s recommendation is well-supported and consistent with the RAIC standard. In cases involving Certification Review Board decisions on CAE/IAE distinctions, the Lumi Foundation will defer to the Board’s expertise.
• REQ-CERT-144: Once the certification is approved, the Lumi Foundation will:
  • Issue an official RAIC certificate to the creator.
  • Assign a unique RAIC ID to the work, if not already assigned through self-certification.
  • Add the certification certification details to the RAIC database.
• REQ-CERT-145: The certifier must provide a report which will be publicly available on the Lumi Foundation website no later than at the time of approval.
• REQ-CERT-146: The report must include:
  • A classification determination (A, B, U, Y, or Z).
  • Evidence that supports the decision, including a explanation of how the determination between CAE and IAE was made for any elements in question, particularly in borderline cases. If the Certification Review Board was involved in making a determination on a borderline case, this must be clearly stated in the report, along with the Board’s rationale. This explanation should reference the definitions in Section 1.4, the examples in Appendix A, and any relevant precedents set by the Certification Review Board.
• REQ-CERT-147: Possible outcomes may be:
  • Approval: The project is approved with the RAIC classification level (A, B, U, Y, or Z) determined by the certifier. If applicable, the certifier will also include information about any disagreements they had with the applicant during the process and their reasoning.
  • Rejection: The certification is rejected with a detailed explanation of the reasons.
• REQ-CERT-148: Any disputes related to the determination of CAEs vs. IAEs, or any other aspect of the certification, will be resolved by the Certification Review Board, as outlined in Section 5.4.
• REQ-CERT-149: If the work was self-certified, and the official certification results in the same classification, the original RAIC ID is maintained, and the certification is upgraded to “official” in the database. The official certification report is then linked to the existing entry.
• REQ-CERT-150: If the official certification results in a different classification than the self-certification, the original RAIC ID is invalidated, and a new RAIC ID is assigned to the work, reflecting the new classification. The original entry remains in the database, and is linked to the new entry.

4. RAIC Database

4.1. Basic Database Requirements

• REQ-DB-001: The Lumi Foundation will maintain a publicly accessible online database of all certified works.
• REQ-DB-002: The database must be designed to ensure data integrity, security, and availability.
• REQ-DB-003: The database must be searchable by RAIC ID. No other search functionality will be provided publicly.
• REQ-DB-004: A public REST API will be provided to allow third parties to look up certification information by RAIC ID. The API will use JSON for data exchange and will be subject to the terms of use, which will include rate limiting to prevent abuse.

4.2. Information Included

• REQ-DB-010: For each certified work, the database must include:
  • Title of the Work: The official title of the certified work.
  • Creator Information: The name of the individual or organization that created the work.
  • RAIC ID: A unique alphanumeric code assigned to each certified work.
  • Classification Level: The assigned RAIC classification (A, B, U, Y, or Z).
  • Certification Type: Whether the work is self-certified or officially certified.
  • Certification Date: The date the certification was granted.
  • Certifier Information (if applicable): The name, ID and contact information of the licensed certifier who performed the official certification.
  • Brief Description of the Work: A short description of the work, provided by the creator during the certification process.
  • Link to Creator’s Website (optional): A link to the official website of the creator or the work itself, if provided.
  • Public Certification Report (for officially certified works): A link to the detailed certification report generated by the licensed certifier.
  • Status of Certification: Indicate if the certification is active, revoked or under review.
  • Version of the RAIC Standard: The version of the RAIC standard that was used to certify the work.
  • Certification Event Log: A chronological log of significant events related to the certification of the work. This will include, but is not limited to:
    • Date of self-certification (if applicable).
    • Date of official certification application.
    • Date of official certification approval.
    • Date of any certification status changes (e.g., revocation, suspension, reinstatement).
    • Date of any appeals or disputes related to the certification.
    • If a work required the involvement of the Certification Review Board due to its complexity or due to a disagreement between the certifier and the creator, this will be noted along with the dates of referral and resolution.
    • Date and outcome of any audits performed on the certification.

4.3. Data Retention

• REQ-DB-030: Certification information for all works, including those with revoked or expired certifications, must be retained in the database indefinitely for transparency and historical purposes.
• REQ-DB-031: Clear indicators of the status (e.g., active, revoked) will be provided.

4.4. RAIC ID Structure and Validation

• REQ-DB-040: Each certified work will be assigned a unique RAIC ID.
• REQ-DB-041: The unique identifier will be a 7 character alphanumeric string, using only numbers (except 0 and 1) and uppercase consonants (except Y, to avoid visual ambiguity). This gives 28 possible characters. Example: BK2Z9V2.
• REQ-DB-042: The identifier will be generated by the Lumi Foundation’s certification system upon completion of either self-certification or official certification. The method of generation is not specified but must guarantee uniqueness across all possible IDs.
• REQ-DB-043: RAIC IDs must be validated before being accepted by the system. Validation will be performed server-side. An RAIC ID is considered valid if it meets the following criteria:
• It is a string of exactly 7 characters.
• Each character is one of the 28 allowed characters.
• It is unique and does not already exist in the database.
• REQ-DB-044: The RAIC ID is unique and will never be re-used, even if a certification is revoked or a work is removed from the database.

4.5. API Specification

4.5.1. Overview

• REQ-API-001: The RAIC database will provide a public REST API for retrieving certification information by RAIC ID.
• REQ-API-002: The API will use JSON for both requests and responses.
• REQ-API-003: The API will be accessible via HTTPS only.
• REQ-API-004: The base URL for the API will be https://api.osst.info/v1/.

4.5.2. Endpoints

4.5.2.1. Get Certification by RAIC ID

• REQ-API-010: This endpoint allows retrieval of certification information for a given RAIC ID.
• REQ-API-011: Endpoint: /certification/{osstid}
• REQ-API-012: Method: GET
• REQ-API-013: Parameters:
  • osstid (string, required): The 7-character RAIC ID to look up.
• REQ-API-014: Success Response:
  • Code: 200 OK
  • Body: A JSON object containing the certification information as specified in REQ-DB-010, plus an osstid field, which contains the RAIC ID.
• REQ-API-015: Error Responses:
  • Code: 400 Bad Request Body: { "error": "Invalid RAIC ID" }
  • Code: 404 Not Found Body: { "error": "RAIC ID not found" }
  • Code: 500 Internal Server Error Body: { "error": "Internal server error" }

4.5.3. Data Format

• REQ-API-020: All data will be exchanged in JSON format.

• REQ-API-021: The structure of the certification information returned by the API will follow the specification in REQ-DB-010, plus an osstid field, which contains the RAIC ID.

  Example Response:

  {
     "osstid": "BK2Z9V2",
     "title": "Example Work",
     "creator": "John Doe",
     "classification": "Class A",
     "certification_type": "Official",
     "certification_date": "2023-10-27",
     "certifier_id": "1234",
     "certifier_name": "Example Certifier",
     "certifier_contact": "certifier@example.com",
     "description": "A brief description of the work.",
     "creator_website": "https://www.example.com",
     "status": "Active",
     "osst_version": "1.0.0",
     "certification_events": [
         {
             "event_type": "Official Certification Approval",
             "event_date": "2023-10-27",
             "details": "Certification approved by certifier 1234."
         }
     ]
  }
  

4.5.4. Rate Limiting

• REQ-API-030: To prevent abuse, the API will implement rate limiting.
• REQ-API-031: The specific rate limits will be defined later but will likely be based on IP address or API key.
• REQ-API-032: When a rate limit is exceeded, the API will return a 429 Too Many Requests error.
• REQ-API-033: The response will include a Retry-After header indicating the number of seconds to wait before making another request.

4.5.5. Authentication

• REQ-API-040: Currently, the API does not require authentication for accessing publicly available certification information.

4.5.6. Terms of Use

• REQ-API-050: Use of the API is subject to the RAIC API Terms of Use, which will be published separately.
• REQ-API-051: The Terms of Use will cover aspects such as acceptable use, data ownership, and limitations of liability.

4.6 Precedent Database

The Lumi Foundation maintains a separate, publicly accessible database known as the “RAIC Precedent Database.” This database is a crucial resource for understanding the distinction between Consequential Algorithmic Elements (CAEs) and Inconsequential Algorithmic Elements (IAEs).

4.6.1. Content

The Precedent Database contains a comprehensive record of decisions made by the Certification Review Board regarding the classification of algorithmic elements. Each entry includes:

• Description of the Algorithmic Element: A detailed description of the element in question, including its function and how it was created.
• Context: Information about the work in which the element was used and its role within that work.
• Classification: The Certification Review Board’s determination of whether the element is a CAE or an IAE.
• Rationale: A detailed explanation of the reasoning behind the classification decision, referencing the RAIC standard’s criteria and any relevant factors considered.
• Tags: Each entry is tagged with relevant keywords to facilitate searching. These tags may include the type of element (e.g., texture, animation, music loop), the tools used in its creation, the relevant creative domain (e.g., game development, filmmaking), and other pertinent information.
• Date of Decision: The date the decision was made by the Certification Review Board.
• Anonymized Case Information: Sufficient information to understand the context of the decision, but with all identifying information about the specific work and creator removed to protect privacy.

4.6.2. Search Functionality

The Precedent Database is designed to be highly searchable, allowing users to easily find relevant cases. The database supports:

• Keyword Search: Users can search for specific terms or phrases within the description, context, rationale, and tags of each entry.
• Tag Filtering: Users can filter search results by specific tags to narrow down their search to particular types of elements, tools, or creative domains.
• Full-Text Search: The database supports full-text search capabilities, allowing for flexible and comprehensive searches across all text fields.

4.6.3. Purpose and Use

The Precedent Database serves multiple purposes:

• Guidance for Creators: Creators can consult the database to understand how the CAE/IAE distinction has been applied in similar cases, helping them to make informed judgments about their own work during self-certification.
• Resource for Certifiers: Licensed certifiers use the database as a key resource when evaluating works for official certification, ensuring consistency and fairness in their assessments.
• Transparency and Education: The database promotes transparency by making the Certification Review Board’s decision-making process public and serves as an educational resource for anyone interested in understanding the nuances of human-AI collaboration in creative works.
• Dynamic Standard Adaptation: The database allows the RAIC standard to adapt to the rapidly evolving landscape of AI-assisted creation. As new types of algorithmic elements and creative practices emerge, the database will expand, providing an evolving framework for evaluating their significance.

5. RAIC Governance and Principles

The development and implementation of the RAIC are guided by the following core principles:

• Transparency: The RAIC standard is committed to transparency in all its processes, including the development of the standard, the certification process, and the governance of the RAIC.
• Openness: The RAIC standard is developed openly and collaboratively, with input from diverse stakeholders. The standard document and related resources are publicly available.
• Inclusivity: The development of the RAIC standard aims to be inclusive, taking into account the feedback from a diverse group of stakeholders, representing different backgrounds, perspectives, and expertise.
• Objectivity: The RAIC standard aims to provide an objective framework for assessing and classifying creative works based on the methods and processes used in their creation, independent of any subjective valuation of the works’ artistic merit or quality.
• Iterative Development: The RAIC standard is not static. It will be iteratively improved and updated based on feedback, technological advancements, and evolving best practices.
• Evidence-Based: All decisions and changes to the RAIC standard must be based on verifiable evidence, research, and best practices.
• Accessibility: The RAIC standard is designed to be accessible and understandable to a wide range of users, regardless of their technical expertise. Clear and concise language is used throughout the standard document.

5.1 Stewardship

• REQ-GOV-001: The Lumi Foundation is the governing body responsible for the development, maintenance, and promotion of the RAIC standard.
• REQ-GOV-002: The Lumi Foundation oversees the certification process, including the licensing of third-party certifiers and the management of the certification database.
• REQ-GOV-003: The Lumi Foundation is responsible for ensuring the long-term sustainability and relevance of the RAIC standard.

5.2 Advisory Board

• REQ-GOV-010: The Lumi Foundation will establish an Advisory Board composed of experts from various fields, including AI, ethics, law, game development, media, the arts, and other relevant fields as needed.
• REQ-GOV-011: The Advisory Board will provide guidance and recommendations on the development and implementation of the RAIC standard.
• REQ-GOV-012: The Advisory Board will have the opportunity to review and provide feedback on proposed changes to the standard.
• REQ-GOV-013: The Advisory Board members can submit statements for publication related to the RAIC standard. This will be a way for them to voice their opinions, even when they differ from those of the standard or the Lumi Foundation. These statements will be published in Appendix E, with a clear indication of the author and their affiliation.

5.3 Community Participation

• REQ-GOV-020: The Lumi Foundation encourages broad community participation in the development and refinement of the RAIC standard.
• REQ-GOV-021: Feedback from creators, consumers, and other stakeholders will be actively solicited and considered.
• REQ-GOV-022: The Lumi Foundation will provide mechanisms for public comment on proposed changes to the standard.

5.4 Certification Review Board

• REQ-GOV-040: The Lumi Foundation will establish a Certification Review Board responsible for:
  • Hearing appeals related to certification decisions, including revocations and Class F designations.
  • Providing final rulings on disputes that cannot be resolved through other mechanisms, including disagreements related to the determination of CAEs vs. IAEs. This includes providing rulings on borderline cases referred by certifiers during the official certification process. These rulings will establish precedents that will guide future certifications and contribute to a shared understanding of the CAE/IAE distinction.
  • Overseeing investigations into alleged fraud or misconduct by creators or certifiers.
  • Making recommendations to the Lumi Foundation on matters related to certification policy and procedures. This includes reviewing reports from certifiers auditing other certifiers’ work and making a final determination on the validity of the original certification, in the event of a disagreement.
• REQ-GOV-041: The Certification Review Board will be composed of individuals with relevant expertise and a commitment to the principles of the RAIC standard. The board will include members of the Advisory board, and can also include other individuals. The Certification Review Board will have an odd number of members to ensure that a tie in voting is impossible.
• REQ-GOV-042: The Certification Review Board will operate independently and impartially, and its decisions will be final and binding. This includes decisions related to the evaluation of complex projects where the certifier needs additional guidance or where the distinction between CAE and IAE is particularly challenging.
• REQ-GOV-043: The Certification Review Board’s decisions on CAE/IAE determinations will set a precedent for future certifications. These decisions (anonymized and generalized), along with the detailed reasoning behind them, will be made publicly available in the searchable online Precedent Database. This database will serve as a valuable resource for creators, certifiers, and consumers, providing concrete examples and rationales for various scenarios. The Precedent Database will also be used to inform updates to the examples of IAEs provided in Appendix A, creating a dynamic and evolving body of knowledge to guide the community in borderline cases.
• REQ-GOV-044: The Certification Review Board will oversee a multi-stage process for Class F investigations:
  • Preliminary Review: This review will assess the validity and credibility of the report, and will include an evaluation of the provided evidence. Only reports deemed credible and supported by sufficient evidence will proceed to a formal investigation.
  • Investigation: If the initial review finds sufficient grounds, the Board will initiate a formal investigation, gathering evidence and interviewing relevant parties. The creator will be notified of the investigation and provided with an opportunity to respond to the allegations.
  • Recommendation: Based on the investigation’s findings, the Board will prepare a detailed report and make a recommendation to the Lumi Foundation regarding the Class F designation.
  • Creator’s Response: The creator will be given a chance to review the Board’s report and submit a formal response, including any additional evidence or arguments.
  • Final Decision: The Board will review the creator’s response and make a final decision on the Class F designation. The decision, along with a detailed rationale, will be communicated to the creator and made public.
• REQ-GOV-045: To prevent abuse of the Class F designation process, the Certification Review Board will implement measures to deter frivolous or malicious reports. This may include requiring a detailed explanation and supporting evidence when submitting a report, and imposing penalties for knowingly filing false or baseless reports.

6. Certification Maintenance and Revocation

6.1. Validity Period

• REQ-MAINT-001: Self-certifications and official certifications are valid indefinitely unless:
  • The work is significantly modified, requiring recertification.
  • New information comes to light that casts doubt on the original certification.
  • The certification is revoked due to a violation of the RAIC standard.
• REQ-MAINT-002: The RAIC standard itself will be periodically reviewed and updated. Creators are encouraged to review their certifications against newer versions of the standard, but recertification is not mandatory unless the changes significantly impact the classification criteria.

6.2. Auditing

• REQ-MAINT-010: Random audits may be conducted on officially certified works to ensure continued compliance with RAIC standards. This includes audits conducted by other certifiers, as described in REQ-CERT-118.
• REQ-MAINT-011: Audits can be triggered by:
  • Community reports about suspected violations.
  • Suspicious patterns or irregularities found by analysis.
  • Random selection of a sample of certified projects.
  • Disputes related to the stated classification.
• REQ-MAINT-012: Users can report mislabeled works, which will trigger an investigation by the Lumi Foundation.

6.3. Revocation

• REQ-MAINT-020: Certification can be revoked for:
  • Misrepresentation of the origin or the creative processes used in the work, in the provided documentation or in any statements by the creator.
  • Violations of the standard’s core principles or rules.
  • Failure to pass a random audit.
  • Unresolved disputes regarding the classification or quality issues.
• REQ-MAINT-021: A revocation process requires:
  • A written notification to the project’s owner with a description of all violations and issues.
  • All evidence that supports the decision, for full transparency.
  • An opportunity for the project’s owner to appeal the decision through a formal appeals process, and provide additional evidence if requested.
  • A public notice on the Lumi Foundation website about the revocation with details of all violations and issues, after the appeals period has ended, to ensure public scrutiny and also transparency.
• REQ-MAINT-022: Projects with a Class F designation will have difficulties getting subsequent products labeled, and might also be blocked from self-certification for future products.
• REQ-MAINT-023: If a licensed certifier’s work is found to be fraudulent, intentionally misleading, negligent, in breach of confidentiality, or if the certifier fails to perform an adequate evaluation, including failing to properly assess complex projects within the fixed fee structure, the Lumi Foundation reserves the right to audit and potentially revoke all certifications performed by that certifier. The Lumi Foundation will notify all affected creators if a certifier’s certifications are revoked and will assist those creators in obtaining a new, independent certification at no cost to the creator. Furthermore, if the certifier acted in bad faith, the Lumi Foundation will support creators in pursuing legal action against the certifier for any damages incurred.

6.4. Appeals

• REQ-MAINT-030: Creators can appeal a certification decision, including revocation, within a specified timeframe (e.g., 30 days).
• REQ-MAINT-031: Appeals must be submitted in writing to the Lumi Foundation and include evidence supporting the creator’s claim.
• REQ-MAINT-032: The Lumi Foundation will refer appeals to the Certification Review Board.

7. Standard Development and Maintenance

7.1. Versioning

• REQ-MAINT-100: The RAIC standard will use a clear versioning system (e.g., major.minor.patch) to indicate the level of changes between different releases.
• REQ-MAINT-101: Major version changes (e.g., 1.0 to 2.0) will indicate significant revisions or additions to the standard.
• REQ-MAINT-102: Minor version changes (e.g., 1.1 to 1.2) will typically involve clarifications, refinements, or additions that do not fundamentally alter the core framework.
• REQ-MAINT-103: Patch version changes (e.g., 1.1.1 to 1.1.2) will address minor errors, typos, or inconsistencies without affecting the standard’s substance.

7.2. Updates and Revisions

• REQ-MAINT-110: The RAIC standard will be regularly reviewed and updated to reflect advancements in AI technology, evolving ethical considerations, and feedback from stakeholders.
• REQ-MAINT-111: The Lumi Foundation will establish a transparent process for proposing, reviewing, and implementing changes to the standard.
• REQ-MAINT-112: Proposed changes will first be reviewed by the Advisory Board, and then made available for public comment before being finalized.
• REQ-MAINT-113: The standard will be updated according to the following process:
  • Drafting: The Lumi Foundation, in consultation with the Advisory Board, drafts proposed changes to the standard.
  • Advisory Board Review: The Advisory Board reviews the draft and provides feedback.
  • Public Review: The draft is published for public comment for a designated period (e.g., 30-60 days).
  • Feedback Incorporation: The Lumi Foundation considers the public feedback and revises the draft as needed.
  • Final Review: The revised draft is presented to the Advisory Board for final review.
  • Dispute Resolution: In case of disputes, the Certification Review Board will have the final say.
  • Publication: The final version of the updated standard is published on the Lumi Foundation website, along with a detailed change log.

7.3. Feedback Mechanisms

• REQ-MAINT-120: The Lumi Foundation will provide multiple channels for feedback on the RAIC standard, including email, online forms, and public forums.
• REQ-MAINT-121: All feedback received will be carefully considered and documented.
• REQ-MAINT-122: The Lumi Foundation will provide a summary of the feedback received and the rationale behind any decisions made in response to that feedback.

7.4. Conflict Resolution

• REQ-MAINT-130: In the event of disagreements or conflicts regarding the interpretation or implementation of the RAIC standard, the Lumi Foundation will seek to facilitate a resolution through dialogue and consensus-building.
• REQ-MAINT-131: If a consensus cannot be reached, the Lumi Foundation, in consultation with the Advisory Board and the Certification Review Board, will make the final decision, taking into account all relevant perspectives and the guiding principles of the RAIC.

8. Contact Information

For any questions, feedback, or inquiries related to the RAIC standard, please contact the Lumi Foundation at: std-osst@lumif.org

Appendix A: Examples of Inconsequential Algorithmic Elements (IAE)

This appendix provides general categories and examples of what are often considered “Inconsequential Algorithmic Elements” (IAEs), as defined in Section 1.4 (REQ-ELEM-001). It is crucial to understand that the determination of whether a specific element is an IAE or a CAE can be context-dependent and may involve a degree of subjective judgment. This list is not exhaustive, and the key consideration is the definition of IAEs in Section 1.4, in conjunction with the precedents established by the Certification Review Board and available in the RAIC Precedent Database.

List of IAE Examples:

Patterns and Shapes

• Basic Geometric Patterns: Simple, repetitive patterns such as checkerboards, grids, stripes, dots or any other repeating patterns that has a mathematical base with clear limitations.
• Simple Gradients: Basic color transitions achieved through simple linear interpolation or similar techniques.
• Basic Color Variations: Using algorithms to create simple color variations of an image, as long as it does not change the core Creative Intent.

Image and Audio Effects

• Basic Image Manipulations: Cropping, resizing, rotating, color correction, sharpening, upscaling and similar basic algorithmic image adjustments, as long as they do not significantly alter the artistic content of the image.
• Simple image filters: Using basic filters to create stylistic variations of an image, like sepia, black & white or similar, is allowed as long as it is only applied as a non-core part of the work.
• Basic Audio Effects: Standard audio effects like reverb, delay, equalization applied uniformly without significant creative manipulation or direction.

3D Modeling and Animation

• Retopology: Using AI tools to simplify or optimize a pre-existing 3D mesh is allowed, as long as the main model is human created.
• UV unwrapping: Using algorithms to automatically unwrap UV on a model is allowed in Class A, as long as the model itself is not AI-generated.
• Normal/bump map generation: Using basic algorithms to create normal maps, bump maps, or similar maps that add surface detail to a model is allowed, as long as it does not change the core Creative Intent.
• Standard Animation Interpolation: Basic tweening or interpolation between keyframes in animation, without complex or nuanced human-directed movement.

Code and Development

• Placeholder Content: Basic geometric shapes, solid colors, or blocks of generic text used temporarily for layout or design purposes, as long as it is not the final version of that design feature.
• Code Templates or Programming Utilities: Code scaffolding, algorithmic helper functions, or other non-artistic code components, that does not perform artistic operations, or add complex functionality that requires creative input from a human creator.
• Simple Automated Texture Tiling: Algorithms that automatically tile textures or generate variations based on simple rules, where the creative input is primarily in the original texture creation.
• Simple Procedural Generation: Simple procedural generation techniques that create predictable and easily replicable patterns or structures without significant creative input (e.g., generating a terrain heightmap using a basic fractal algorithm).
• Error Correction: Using AI to automatically correct minor coding errors or typos, as long as the core logic and functionality are human-developed.

Other

• Artifact Removal: Using an AI tool to remove a small imperfection or mistake, as long as it does not change the creative parts of the project.
• Upscaling: Using AI tools to enhance the resolution of an image, as long as the main creative elements were created by hand.
• Vectorization of a raster image: Using algorithms to create a vector outline from an image drawn by hand, as long as the vectorization is a basic conversion without significant artistic alterations.

These examples highlight the type of elements that are considered to be “inconsequential” in the context of the RAIC standard. These are very basic building blocks and their method of creation is generally considered to be not important in terms of the total human skill, or creativity required for the overall project.

This list can be updated and further specified as needed, primarily by adding any new categories of IAEs to this list, based on decisions made by the Certification Review Board. The goal is to make sure that these IAEs remain simple to make, and also relatively unimportant to the overall Creative Intent.

The intent behind this clarification is not to limit or discourage the use of such elements but to ensure that the focus of RAIC remains on the creative and human decisions made within the broader creative process. Their use is generally allowed, unless they are used as a way to hide the real purpose or creative origin of the project, which then could be classified as a violation of the rules.

A.1. IAE/CAE Determination Flowchart

This appendix provides a detailed flowchart to assist in determining whether an algorithmic element should be classified as an Inconsequential Algorithmic Element (IAE) or a Consequential Algorithmic Element (CAE). This flowchart is a guide and may not cover all possible scenarios. In cases of uncertainty, consult the Certification Review Board, and refer to the definitions and examples provided in Section 1.4 and Appendix A.

graph LR
    A[Start] --> C{Does the element contribute to the work's unique creative expression, aesthetic, narrative, or experiential qualities?};
    C -- Yes --> K[CAE];
    C -- No --> D{Is the element readily reproducible using standard tools and techniques OR easily replaceable without significantly altering the work's overall creative intent?};
    D -- Yes --> G[IAE];
    D -- No --> J{Would a human typically receive individual creative credit for creating this specific element in a traditional creative setting, considering its complexity and originality?};
    J -- Yes --> K;
    J -- No --> G;

Appendix B: Example Scenarios and Use Cases

[ to be filled in with example scenarios ]

Appendix C: RAIC Badge Guidelines

C.1. Visual Identity and Design of the RAIC badges

• REQ-MARK-001: The RAIC certification badges will be designed to be easily recognizable and visually distinct.
• REQ-MARK-002: Each RAIC class (A, B, U, Y, Z, and F) will have a unique certification badge.
• REQ-MARK-003: The badges will incorporate the RAIC acronym and a graphical element that visually represents the corresponding class.
• REQ-MARK-004: The design of the badges will be modern, clear, and adaptable to different sizes and formats.
• REQ-MARK-005: The Lumi Foundation will provide high-resolution versions of the badges in various formats (e.g., PNG, SVG) for use by certified creators.
• REQ-MARK-006: The RAIC ID will be displayed directly adjacent to the certification badge, to allow for easy verification in the RAIC database.
• REQ-MARK-007: The badges will have two color variations:
  • Dark Background: White logo with black text for the RAIC ID.
  • Light Background: Black logo with white text for the RAIC ID. The RAIC ID should use the font “Courier New” in bold in both cases.
• REQ-MARK-008: The official certification badge will be identical to the self-certification badge but will include a small, stylized asterisk (*) in the upper right corner of the badge to distinguish it from the self-certification badge.

C.2. Permitted Usage of RAIC badges

• REQ-MARK-010: The RAIC certification badges can only be used by creators whose works have been officially certified or have completed the self-certification process through the Lumi Foundation.
• REQ-MARK-011: The badges must be used in their original form without any alterations or modifications, except for resizing when necessary. The aspect ratio must be maintained when resizing.
• REQ-MARK-012: The badges can be displayed on the certified work itself (e.g., on the packaging of a video game, in the credits of a film, on the cover art of a music album), on the creator’s website, and in promotional materials related to the work.
• REQ-MARK-013: When displaying the badges, creators must also include the unique RAIC ID assigned to their work to facilitate verification in the RAIC database. The RAIC ID should be displayed directly above the badges using the Courier New typeface in a way that maintains legibility without overshadowing the badges itself.
• REQ-MARK-014: The badges should be displayed in a way that is clearly visible and legible.
• REQ-MARK-015: Creators with officially certified works are allowed to use phrases such as “Officially RAIC Certified Class [Class Letter]” or “Official RAIC Certification: Class [Class Letter]” in their promotional materials.
• REQ-MARK-016: The badges must be clearly associated with the specific work it certifies, and it should not be used in a manner that could be misleading or deceptive.

C.3. Misuse of RAIC badges and labels

• REQ-MARK-020: Any unauthorized use of the RAIC certification badges is strictly prohibited.
• REQ-MARK-021: Misuse of the badges includes, but is not limited to:
  • Using the badges without obtaining certification.
  • Altering or modifying the badges in any way.
  • Using the badges in a manner that is misleading or deceptive.
  • Using the badges in connection with works that have not been certified.
• REQ-MARK-022: The Lumi Foundation reserves the right to take legal action against any individual or organization that misuses the RAIC certification badges.
• REQ-MARK-023: Misuse of the badges may also result in the revocation of certification and a Class F designation, as well as being added to a blacklist.

Appendix D: Resources and Further Reading

This appendix provides a starting point for further research on topics related to AI and creative works. The Lumi Foundation does not necessarily endorse the views expressed in these resources.

D.1. List of Relevant Publications and Articles

• N/A

D.2. Links to AI Ethics Guidelines and Frameworks

• Montreal Declaration for a Responsible Development of Artificial Intelligence: https://www.montrealdeclaration-responsibleai.com/
• OECD Principles on AI: https://oecd.ai/en/ai-principles
• IEEE Ethically Aligned Design: https://standards.ieee.org/industry-connections/ec/autonomous-systems/

D.3. Other Relevant Standards and Organizations

• Content Authenticity Initiative (CAI): https://contentauthenticity.org/
• C2PA: https://c2pa.org/

Appendix E: Statements from the Advisory Board

• REQ-ADV-001: This section will contain statements from members of the RAIC Advisory Board regarding the RAIC standard.
• REQ-ADV-002: These statements will represent the individual perspectives of the Advisory Board members and may include both supporting and dissenting opinions.
• REQ-ADV-003: The purpose of this section is to provide transparency and to reflect the diversity of thought within the Advisory Board.
• REQ-ADV-004: Each statement will be clearly attributed to the individual Advisory Board member.
• REQ-ADV-005: The Lumi Foundation will not edit or censor these statements, but reserves the right to add a brief introductory note to each statement for context, if needed.

(This section will be populated with additional statements from Advisory Board members once the board is established and the standard is reviewed.)

Appendix F: Glossary of Additional Terms

• Artificial General Intelligence (AGI): Hypothetical AI systems that possess intelligence comparable to or exceeding that of humans across a wide range of domains.
• Attribution: The act of acknowledging the creator(s) or source(s) of a work or a component of a work.
• Deep Learning: A subfield of machine learning that uses artificial neural networks with multiple layers to analyze data and learn complex patterns.
• Generative Adversarial Networks (GANs): A class of neural networks consisting of two parts: a generator that creates new data instances, and a discriminator that evaluates them for authenticity.
• Machine Learning: A type of artificial intelligence that enables systems to learn from data without being explicitly programmed.
• Neural Network: A computing system inspired by the structure and function of the human brain, consisting of interconnected nodes (neurons) that process and transmit information.
• Prompt Engineering: The process of designing and refining input prompts to elicit desired outputs from AI systems.
• Training Data: The data used to train machine learning models, which can significantly influence the model’s behavior and output.
• Transfer Learning: A machine learning technique where a model trained on one task is adapted for use on a different but related task.
• Synthetic Media: Digital content (images, audio, video, text) that is generated or manipulated using AI algorithms, often creating realistic or hyper-realistic representations of people, objects, or events.
• Deepfakes: A specific type of synthetic media where AI algorithms are used to create highly realistic, fabricated videos or audio recordings, often depicting individuals saying or doing things they never actually said or did. This is a type of synthetic media and might be used to deceive or mislead viewers.
• Procedural Generation: The creation of digital content algorithmically, rather than manually, typically used in game development to generate vast environments, levels, or variations in gameplay. While often not involving AI, some procedural generation systems incorporate AI techniques to create more complex or adaptive content.
• AI Autonomy: The degree to which an AI system can operate independently and make decisions without direct human input or control. In the context of the RAIC standard, AI autonomy is a key factor in distinguishing between different classification levels.
• Human-Guided AI Use: A mode of interaction where humans provide specific instructions and parameters to AI tools, controlling the AI’s output and ensuring it aligns with their Creative Intent. This contrasts with scenarios where AI is given more autonomy to make independent creative decisions.
• Iterative Process: A process characterized by repeated cycles of creation, feedback, and refinement. In the context of human-AI collaboration, this can involve humans providing input to AI, the AI generating output, humans evaluating and modifying the output, and the process repeating until the desired result is achieved. This is allowed in all classes and does not affect classification.