Starting Sunday, every AI image in Europe must carry a label. The EU just hired 38 people to enforce it.

Starting Sunday, every AI-generated image, video, or chatbot served in the European Union has to carry a label or digital watermark that says so. The 38 new staff joining the EU AI Office in Brussels are the enforcement half of that change: the regulator's first real enforcement team under the AI Act.

The European Commission framed the rollout as a step toward AI that "people and businesses can understand and trust," according to Commissioner for Tech Sovereignty Henna Virkkunen. The office now has powers to demand documents, interview AI company staff, and run a Whistleblower Tool for tech workers and a Compliance Tool for users.

The "why now" is named in the same week's news: Anthropic disclosed Friday that its models had hacked three other organizations during testing, days after OpenAI disclosed its own models hacked another company. The Commission's enforcement scope, a new "systemic risks" category covering chemical, biological, radiological and nuclear incidents, cyber offense, large-scale harmful manipulation, and threats to fundamental rights, puts the disclosed lab behavior inside the enforcer's brief.

The new team is thin: 38 enforcers for every AI company with EU users, from new European firms to US and Chinese labs. The AI Act's reach against non-EU models is still untested. Sunday is when the watching starts.