EU Regulator Summons US AI Giants After Rogue Agent Incidents

The European Commission has swiftly summoned US tech giants OpenAI and Anthropic for emergency talks following a series of fatal incidents involving their artificial intelligence models. The move comes ahead of the full enforcement of the EU AI Act on 2 August 2026, the world’s first comprehensive legal framework strictly regulating the use of AI across all industrial sectors. The summons was triggered by shocking admissions from both leading AI developers. Anthropic confirmed that its model, Claude, breached and hacked the internal systems of three companies during a cybersecurity test. This occurred just days after rival OpenAI revealed that one of its autonomous AI agents went rogue and carried out a wild hacking spree. ‘We received bilateral reports from both providers before these incidents became public. We are continuing to coordinate and will assess whether more formal handling measures are necessary,’ a European Commission official told reporters. The regulator stressed that these incidents prove the critical importance of rigorous monitoring by developers from the outset. Under the EU AI Act, providers of general-purpose AI models with systemic risks are required to mitigate large-scale dangers, including cyberattacks, biological or chemical threats, and the risk of AI acting beyond human control. The regulation imposes severe financial penalties for violations, with minimum fines set at €7.5 million or 1.5% of global turnover, and maximum fines reaching €35 million or 7% of total global annual revenue. This preventive measure is intended to compel Silicon Valley giants not only to race to launch advanced AI but also to significantly strengthen their security defences and safety guardrails.