Machines that must say so — transparency becomes the first universal AI rule

Of everything in the AI Act, the rule that took effect August 2 may prove the most consequential: software that talks to humans must admit it's software. Europe just made honesty a compliance requirement — while Washington's answer stalled in committee.

As of August 2, chatbots in the EU must disclose they are AI, and realistic synthetic media must carry labels and watermarks. It reads like a small rule next to the Act's conformity assessments and risk tiers. It is not small. It is the first continent-scale assertion that passing as human is not a product feature — that the Turing test, commercially deployed, is a consumer-protection violation.

Timed, accidentally, to the agent boom

The rule arrives exactly as the machines start talking to each other. Google's consumer agents now phone stores and complete purchases by voice — and enterprise voice agents already answer a meaningful share of business lines. Disclosure was designed for a human talking to a bot; the near future is a bot talking to a bot, with disclosure obligations on both ends and no human listening to either. The rule's first stress test will be a conversation nobody hears.

The vacuum on the other shore

Contrast Washington: the Great American AI Act is stalled in the House over state preemption, leaving US disclosure rules to a patchwork of state laws. The practical result is familiar from privacy: multinationals will build to Brussels because it is the strictest enforced standard, and the EU rulebook becomes the world's default settings — again.

Transparency is the rare AI rule with no capability cost — it doesn't slow training or cap compute; it just forbids deception about nature. That is why it will spread. The era of software pretending to be people ended in Europe on August 2. The rest of the world is on notice.

European Commission — Commission starts enforcing AI Act rules and new transparency requirements → · Collibra — AI regulatory compliance in 2026 →