EU discusses agent-security incidents with OpenAI and Anthropic

reports say that both companies informed the Commission about recent AI-agent security incidents and that officials were in contact with them. The Commission said monitoring high-risk systems is necessary. Public information does not establish that the incidents were identical or that an enforcement finding has been made.

Impact: voluntary incident disclosure is becoming part of regulatory readiness. Enterprises should maintain evidence that links agent actions, tool permissions, containment controls and human escalation.