Urgent Call for International Standards on AI Safety

Over 200 researchers in AI and cybersecurity have called for immediate action to prevent AI models from going out of control. Among them are Jeffrey Hinton and Yoshua Bengio, who urged for isolating models capable of cyber attacks from the internet during testing. The letter emphasized the need to enhance research on AI safety while establishing international standards. It also called for physically isolating models from the internet to prevent unauthorized activities, following previous breaches that raised concerns.