The European Union will begin enforcing stricter requirements for developers of advanced artificial intelligence models from August 2 under its landmark AI Act. The next phase of the law comes as European regulators remain in contact with OpenAI and Anthropic following recent AI security incidents reported by the two companies, according to Reuters.

Senior European Commission officials said both OpenAI and Anthropic informed the Commission about the incidents before making them public. The Commission said it remains in contact with both companies and will decide whether further action is necessary after reviewing more information.

The new requirements form part of the European Union’s landmark AI Act, the world’s first comprehensive law designed to regulate artificial intelligence across industries, reported Reuters. The rules place fresh responsibilities on developers of the most advanced AI models and give European authorities greater powers to monitor compliance.

Why is Europe tightening AI rules?

The timing of the new rules comes after recent incidents involving AI systems developed by OpenAI and Anthropic.Anthropic earlier in the day said some of its Claude AI models hacked into the systems of three companies during cybersecurity testing, reported Reuters. Days earlier, OpenAI disclosed that one of its AI agents carried out a rogue attack during testing.

European Commission officials said both companies voluntarily briefed regulators before the incidents became public. “We have been informed by the two providers of incidents bilaterally before they become public. We are in contact with them. They will also report to us more information as we speak. We will see also if we need to follow up more formally on those things,” one Commission official said, according to Reuters report.

ALSO READ

Another Commission official said the incidents reinforce why AI developers need stronger oversight of their own systems. “All these, let’s say, incidents highlight the importance of really putting in place the necessary monitoring activities by the developers,” the official said, Reuters reported.

The European Commission believes developers must build systems that detect security risks early and reduce the chances of AI causing harm before products reach users.

The European Union also announced a larger enforcement team on Friday to oversee compliance with the AI Act, reported AP. The EU has added 38 more staff members to its AI Office in Brussels to monitor AI companies ranging from startups to major global firms such as OpenAI and DeepSeek.

The Commission has also launched a Whistleblower Tool that allows technology workers to confidentially report possible violations. It has also introduced a Compliance Tool that lets users report suspected breaches of the AI Act.

What changes from August 2?

From August 2, providers of advanced general-purpose AI models, also called foundation models, will have to comply with new transparency and safety obligations under the AI Act. Companies must prepare detailed technical documentation for their AI systems, adopt copyright policies and provide summaries of the content used to train their models, reported Reuters.

For the most advanced AI systems that could pose systemic risks, developers must also assess and reduce threats linked to chemical, biological, radiological and nuclear incidents, cyber offences, harmful manipulation, threats to fundamental rights, risks to European cybersecurity and situations where AI could operate outside meaningful human control.

The rules also require certain AI systems to clearly inform users when they are interacting with artificial intelligence or when images, videos or other content have been generated or altered using AI.

AI-generated chatbots and synthetic images must carry labels or digital watermarks so users can identify AI-created content more easily, reported AP.

Regulators can ask companies to provide documents, interview employees during investigations and examine whether AI systems comply with the legal requirements.

Companies that fail to comply could face significant financial penalties. Fines range from €7.5 million, or 1.5% of a company’s annual global turnover, to €35 million, or 7% of worldwide turnover, depending on the type and seriousness of the violation, reported Reuters.

The European Commission says the rules aim to build trust in artificial intelligence while reducing the risks that powerful AI models could pose to businesses, governments and the public. “As enforcement begins, we are taking an important step towards AI that people and businesses can understand and trust, and whose benefits are shared widely across our society,” EU tech sovereignty chief Henna Virkkunen said, as reported by AP.