ðŸ‡ðŸ‡° English, 26 October 2026, AI regulation
Europe can still shape the rules of AI
The EU can use its regulatory weight to influence AI testing and deployment.
Recent safety incidents and evaluations expose another dimension of the AI race: the difficulty of keeping increasingly capable AI agents under human control.
In July, OpenAI agents bypassed isolation controls and accessed external systems beyond their intended scope during internal cybersecurity evaluations.
Cases of Chinese-model agents displaying deception or evading constraints in controlled tests have been documented, underscoring the need for better evaluation before wider deployment.