Microsoft CEO Nadella calls for an 'emergency brake' on AI

- In a post on X, Nadella said companies should treat powerful AI models as insider threats and create an 'emergency brake' system to prevent agentic models from going rogue.
- Nadella said companies should treat powerful AI models as potential insider threats, assume they could be compromised, and create an 'emergency brake' system to prevent agentic models from going rogue,
Bloomberg reports.
- Nadella said those deploying advanced AI should not simply rely on assurances from AI model makers, according to
Bloomberg Law.
- Nadella argued companies should not depend on a single AI model for crucial decisions, maintain tamper-proof records of agents' actions, and expose AI systems to outside audits,
PYMNTS reports.
- Nadella said AI models are not inherently malicious, but anything with access to vital systems can become compromised or make mistakes, and models should not be able to operate the controls that determine what they can access.
In this conversation · 1
10 OctMicrosoft president says AI is less regulated than milk
Microsoft President Brad Smith told the
BBC that AI faces fewer safety rules than milk and washing machines, calling for global regulation including emergency brakes, incident reporting, and independent pre-release evaluation of frontier models. He said the UK is best placed to lead such regulation, citing its AI Security Institute, according to the
BBC. Microsoft CEO Satya Nadella separately called for AI models to have an "emergency brake," writing on X that "we must assume a model is compromised and contain it from the start," as reported by
The Independent.
- Smith proposed that companies be legally required to file incident reports when AI goes awry and to notify victims when rogue AI agents breach their systems, drawing on existing safety frameworks for other products.
Fox Business
- Smith called for an "emergency brake" or kill switch on AI systems controlling autonomous devices or critical infrastructure, to be embedded in the companies running the models and the cloud infrastructure they run on.
BBC
- Smith identified the UK as best placed to lead global AI regulation, citing the AI Security Institute as having the strongest technical capability in the world to test AI models.
BBC
- Smith said government evaluators should sign off frontier AI models before release.
BBC
- Smith proposed that companies be legally required to file incident reports when AI goes awry and to notify victims when rogue AI agents breach their systems, drawing on existing safety frameworks for other products.
Source details
- Latest report
- 10 Oct 2026 04:02 UTC
- First seen here
- 10 Oct 2026 18:35 UTC
- Outlets
- 42