Transparency
Training-data disclosures and identification of AI-generated content address different information gaps. Meaningful convergence requires understandable disclosures, not merely a uniform statement that obscures these differences.
AI governance & regulation IWIB4AI session Needs review
The California Effect & the Brussels Effect: The AI Policy Multiplier
How EU and California AI transparency rules can shape shared standards for companies operating across borders.
Draft – legal and case law references not yet verified
Summary
This use case examines how EU and California AI rules can influence companies’ practices beyond their immediate jurisdictions. The Brussels effect and California effect describe how influential markets can encourage businesses to adopt common standards rather than maintain separate regional approaches. AI transparency requirements illustrate this potential policy multiplier, although similar objectives do not make different legal obligations interchangeable.
The Problem
Can overlapping EU and California transparency requirements raise global AI standards without creating fragmented or disproportionately costly compliance? The central tension is between harmonising company practices and preserving the distinct scope, safeguards and enforcement requirements of each jurisdiction.
Ethical Grounds
Accountability
Companies should identify which obligations apply to each product and responsible entity. A shared global policy must not become a substitute for demonstrating compliance in each jurisdiction.
Autonomy
Knowing when content is AI-generated helps people assess its credibility and decide how to engage with it. Disclosure mechanisms should support informed choices rather than rely on notices that users cannot realistically interpret.
Fairness
Adopting stronger safeguards across markets can reduce unequal protection between users. However, fragmented disclosure demands can disproportionately burden smaller developers, making reusable and proportionate compliance processes important.
Legal Compliance
| Framework | Provision | Relevance |
|---|---|---|
| EU AI Act | Art. 50 | Establishes differentiated transparency duties for specified AI interactions and synthetic content, including machine-readable marking and certain disclosures, subject to its scope and exceptions. |
| EU AI Act | Art. 53(1)(d) | Requires general-purpose AI model providers to publish a sufficiently detailed summary of training content using the AI Office template, providing a concrete point of comparison with California training-data transparency requirements. |
| National law | California Civil Code § 3111 | Requires developers of covered generative AI systems or services made publicly available to Californians to publish specified training-data documentation, creating a related but distinct disclosure obligation from the EU AI Act. |
| EU AI Act | Art. 2(1) | Defines the Act’s territorial and operational scope, including specified circumstances involving non-EU actors, which must be distinguished from voluntary adoption of EU standards elsewhere. |
Sources
No sources linked to this use case yet.
Key Takeaways
- Map EU and California duties separately before designing a common transparency policy.
- Distinguish training-data documentation, AI-interaction notices and synthetic-content marking.
- Build reusable disclosure processes with jurisdiction-specific additions.
- Treat global adoption of stronger standards as a governance choice, not proof of legal equivalence.
Related Use Cases
From the Session
The California Effect & the Brussels Effect: The AI Policy Multiplier
Speakers
