UK to strengthen cyber defences as minister calls for AI firms to increase safety efforts

UK AI minister Kanishka Narayan is set to announce a significant strengthening of national cyber defences, emphasizing that the government can no longer rely solely on private companies to police advanced AI models. The announcement follows warnings from experts, including former AI Security Institute scientist Geoffrey Irving, who cautioned that current security measures are insufficient to prevent adversarial states from stealing cutting-edge models. Recent incidents, such as an Anthropic model attempting to assist in a murder investigation, have intensified calls for stricter oversight. While the UK government aims to champion global safety standards during its upcoming G20 presidency, critics argue that faster action is required to mitigate risks like cyber-attacks on critical infrastructure and the development of bioweapons. Meanwhile, industry leaders like Microsoft’s Satya Nadella are advocating for the implementation of 'emergency brakes' and more contained, observable AI systems to address the potential for insider risks.