Policy and law

Transparency obligations. Providers of generative systems must mark output in a machine-readable way; deployers must disclose deepfakes and AI-written text published on matters of public interest. Applies from August 2, 2026, with a transition period for systems already on the market.

The voluntary route to Article 50 compliance. Final text published June 10, 2026; the Commission found it adequate on July 8, 2026. Signatories commit to a two-layer approach, secured metadata plus watermarks, and to keeping marks non-removable to the best of their ability.

The Official Journal version of the AI Act, for the exact wording of Article 50 and the definitions it relies on.

AB 853 was signed October 13, 2025. Covered providers owe disclosures and detection tools from August 2, 2026. Large online platforms and generative AI hosting platforms are covered from January 1, 2027, including a ban on knowingly stripping provenance data. Capture device makers follow on January 1, 2028.

The rules for anyone building on Claude, including the deceptive-use section that covers passing generated text off as human work. Read it before building a rewriting tool on the API.

The equivalent rules for OpenAI models, with the same restrictions on deception and academic dishonesty.

Why you do not see humanizers advertised on Google. The policy bars ads for academic-dishonesty services, so the category competes on search rankings and affiliate reviews instead.

The press release that opened the current round of the arms race, with the scope and language limits in its footnotes.