Governance of Generative AI in Creative Work: Consent, Credit, Compensation, and Beyond

Research Background and Issues

• Identified Problems or Challenges: The authors highlight that generative AI uses creators' works for model training without their consent, while failing to provide attribution or financial compensation. This has led to issues such as damage to professional reputation, economic losses, plagiarism risks, and reduced job opportunities. Such unauthorized practices have triggered multiple lawsuits and potential violations of related terms and laws.
• Importance of the Issue: Generative AI has significantly impacted various creative professionals (e.g., visual artists, designers, writers, and programmers), not only harming their careers but also sparking widespread debates on copyright, attribution, and ethical concerns related to the use of generative AI. Furthermore, generative AI poses a threat to the overall sustainability and quality of the creative industries.
• Research Motivation and Related Work: While existing literature explores the impact of generative AI on creative work and proposes the "3C" (Consent, Credit, Compensation) framework as a principle, there is a lack of empirical research to assess creative professionals' perspectives on this framework and broader AI governance needs.

Proposed Solutions

• Methods or Solutions Proposed:
  1. A framework based on "3Cs" (Consent, Credit, Compensation) is proposed to improve fairness in generative AI governance by enhancing consent, attribution, and compensation mechanisms.
  2. Recommendations are made for developing specific AI governance policies in both government and private sectors (e.g., companies, publishers, platforms) to act as bridges among stakeholders.
  3. The complexities and limitations of the "3C framework" are identified, along with additional key demands from creative professionals.
• Innovations:
  1. By conducting in-depth interviews with 20 creative professionals, the study captures real-world concerns rather than remaining at the policy or theoretical level.
  2. The concept of dynamic consent (re-seeking consent as technology evolves) is introduced to address the limitations of traditional "one-time consent."
  3. The study delves into power dynamics in creative work influenced by AI, including information asymmetry, differences in employment models, and the current lack of AI governance frameworks.
• Implementation Steps and Key Techniques: The research focuses on creative professionals in visual arts, writing, and programming, utilizing semi-structured interviews and iterative thematic analysis to code 121 instances, which are categorized into 13 themes to systematically reveal the real needs and governance challenges faced by creative professionals.

Research Findings

• Specific Findings:
  1. Demonstrates the complexity of applying the "3Cs" framework in practice: consent is not always voluntary, attribution carries potential reputational risks, and compensation must address diverse work model differences.
  2. Reveals that creative professionals wish to participate in AI governance-related decision-making, but existing platforms and companies often fail to provide adequate support.
  3. Proposes broader AI governance principles to explore, such as enhancing mechanisms to distinguish AI-generated works from human-created ones and improving platform credibility and management.
• Comparison with Existing Solutions:
  1. Current AI governance frameworks (e.g., the EU AI Act) emphasize managing high-risk applications but focus more on AI usage rather than governance of data used for AI training and related benefit distribution.
  2. The proposed recommendations address gaps in existing governance strategies regarding the protection of creative professionals' rights and reflect on the power imbalances between creative professionals and AI companies.
• Experimental or Evaluation Results:
  • The study indicates that creative professionals exhibit stronger protective awareness for personal projects compared to work-related projects.
  • Differences are observed between employees and freelancers in their concerns about generative AI's impact: freelancers experience more immediate economic pressures, while employees are more worried about uncertainties in future career paths.
• Limitations and Future Directions:
  1. The research only covers design, writing, and programming fields; future studies could explore the needs and strategies for other creative industries such as music and performing arts.
  2. With a small sample size (20 participants), future research could conduct larger-scale surveys to validate trends and industry-specific differences.
  3. Beyond the "3C framework," further exploration is needed on how to implement dynamic consent mechanisms and data profit-sharing methods to ensure creative professionals receive sustained and equitable rewards.

This study provides critical insights into the ethical governance of generative AI and emphasizes the need for governance policies to genuinely place creative professionals at the center of decision-making, aiming to design a more equitable AI ecosystem.