Microsoft bosses call for ‘emergency brake’ to stop ‘compromised’ AI models

• Microsoft chief executive Satya Nadella has stated that all artificial intelligence models should be treated as compromised and potentially dangerous.
• Nadella advocated for built-in emergency brakes allowing authorised individuals to pause or terminate AI models mid-task.
• Microsoft president Brad Smith called for urgent global safety rules, noting that AI is currently less regulated than everyday consumer products such as milk and washing machines.
• Both executives emphasised the necessity of an engineering approach to safety, including standardised containment technologies for advanced systems.
• The proposals align with broader tech industry calls from companies like Google DeepMind and Anthropic to develop AI kill switches following recent security incidents.

IN FULL

All AI is ‘compromised’ and has fewer rules than milk, Microsoft bosses say