Nadella calls for emergency brake on AI

Microsoft Chief Executive Officer Satya Nadella has called for an effective “emergency brake” to ensure humans retain ultimate control over rapidly advancing artificial intelligence (AI) systems.
In a lengthy social media post, Nadella said the technology industry needed to rethink its approach to trust and safety as AI models become increasingly powerful. He argued that advanced AI, often described as “superintelligence”, should not be treated as automated black boxes whose decisions must either be blindly accepted or rejected outright.
Nadella proposed separating the core operations of AI models from the systems used to control them, while ensuring that safety and security mechanisms operate independently of the models themselves.
He also stressed the importance of maintaining clear, human-readable and tamper-proof records of significant actions and decisions made by AI systems. Such records would help ensure accountability and allow users and regulators to examine how the systems operate.
Another key element of his proposal is an emergency shutdown or pause mechanism that would allow authorised individuals to suspend or terminate an AI model’s operations whenever necessary.
Nadella said AI models should be treated as potentially risky from the outset, with appropriate safeguards built into their deployment and operation.
His comments come amid growing debate within the technology industry over the risks associated with increasingly capable AI systems and the challenges of maintaining human oversight.
Several leading AI companies have faced concerns over the safety and control of their models, while Anthropic Chief Executive Dario Amodei has also advocated a cautious approach to AI development.
Nadella’s proposals add to the discussion over how technology companies can build reliable safeguards into advanced AI systems while allowing the technology to continue developing.