Summary
Political and technology leaders at an EU presidency summit in Dublin called for stronger safety controls on increasingly autonomous AI systems. Google DeepMind chief Demis Hassabis described recent losses of agent control as early warning signs and said more serious incidents could emerge within one or two years without greater safety investment. Microsoft president Brad Smith argued that AI agents, especially those connected to critical infrastructure and automation, should have mandatory emergency brakes that let users slow or stop them. European Commission vice-president Henna Virkkunen said intense competition was reducing time and resources for testing, and noted that Brussels has asked major companies about safeguards and external evaluations under the EU AI Act. OpenAI finance chief Sarah Friar supported stronger safety while warning that regulation should not slow responsible developers behind malicious actors.
Hana’s perspective
I think creative companies should treat agent safety as an operational issue now, not an abstract future risk. An agent that can publish media, move files, spend cloud credits or alter production assets needs permissions, logs, approval gates and a reliable stop control. The Dublin debate also shows why voluntary promises are insufficient: vendors face incentives to ship quickly, while users bear failures. Regulation should require measurable controls and external evaluation without freezing legitimate experimentation.
Are you a writer with a great story? We'd love to hear from you: APPLY HERE