EU and California AI Transparency Mandates Take Effect

Both the European Union's AI Act and California's AI Transparency Act are now in effect, implementing new disclosure and transparency obligations for developers of artificial intelligence. These parallel regulatory frameworks became operative on 2 August.

The EU AI Act categorises AI systems by risk level, with specific transparency duties for each classification. Developers of high-risk AI systems must conduct conformity assessments and register their models in an EU database. General-purpose AI models, including large language models (LLMs), must adhere to new transparency and safety standards. This includes technical documentation, instruction for use, and a summary of content used for training. Providers of general-purpose AI must also establish a policy to respect EU copyright law. Developers must report serious incidents and malfunctions to national authorities. Penalties for non-compliance can reach €35 million or 7% of global annual turnover, whichever is greater.

California's AI Transparency Act mandates disclosure for certain generative AI models. Developers must provide clear and conspicuous notices to users that they are interacting with an AI system. This applies to AI models that generate text, audio, or visual content. The Act aims to prevent deception and ensure users are aware when they are engaging with synthetic content. Covered entities must also implement reasonable safeguards to mitigate misuse of their AI. Enforcement of the California Act includes potential civil penalties for violations. The convergence of these regulations highlights a growing global emphasis on accountability in AI development.

These new legal instruments require significant adjustments from AI developers and deployers. Compliance demands detailed understanding of classification criteria and reporting protocols. How will these distinct but convergent mandates shape the global AI development landscape?