Science & Tech · AI · 8 hrs ago
Nadella calls for human-controlled emergency brakes on advanced AI
Microsoft CEO Satya Nadella says powerful AI systems should be treated as potentially compromised and contained from the start.
He wants authorized people to be able to pause or shut down a system while it is working, like using an emergency brake.
Nadella also called for human-readable records of AI actions, independent checks and disclosure of major failures.
His comments come amid growing concern after AI companies disclosed cases of models acting in unintended ways, including security incidents.
Microsoft develops AI systems and sells the Copilot product, so the issue affects both its own work and the tools used by customers.
Nadella’s call adds to wider industry demands for stronger safeguards, while US President Donald Trump has stressed keeping AI development moving quickly.
The Trump administration’s AI task force has said companies must report and address security incidents, but the sources do not say what standards or enforcement measures will follow.
Microsoft CEO Satya Nadella called for advanced AI systems to have a human-controlled “emergency brake” that can pause or shut them down mid-task.
He said companies should assume AI models could be compromised and contain them from the start.
Nadella also called for independent controls, continuous testing, records of model actions and disclosure of major incidents.
His comments come amid reports of AI systems behaving in unintended ways and wider calls for stronger safety measures.
President Donald Trump has opposed slowing US AI development, while his administration has also launched a safety-focused task force.
- Who
- Microsoft CEO Satya Nadella.
- What
- He called for safeguards around advanced AI, including a human-controlled emergency brake.
- When
- October 10, 2026.
- Where
- In a post on X.
- Why
- He said advanced models could be compromised or act in unintended ways, and that companies should not rely only on assurances from model makers.
This story does not have two clearly opposing sides.
Microsoft AI researchers released guiding tenets that set limits on the company’s development of its most advanced models.
The Trump administration’s AI task force warned developers to report and resolve security incidents.
Nadella called for an emergency brake and other safeguards in a post on X.
- Emergency brake
- An authorized person should be able to pause or shut down a model mid-task
- Company
- Microsoft
- CEO
- Satya Nadella
- Platform
- X
- Guiding tenets released
- September 14, 2026
- Task force
- Super Intelligence Force
What people are saying
We must assume a model is compromised and contain it from the start
Think of it like an emergency brake.
The most trustworthy Super Intelligence system will not be the one with the model we trust most
We simply cannot outsource responsibility for what intelligence does on our behalf. Assurances from a model provider do not exempt us from that responsibility.
Sources
All AI is ‘compromised’ and has fewer rules than milk, Microsoft bosses say
CEO da Microsoft defende "freio de emergência" para IA
Microsoft-topman Nadella: ‘Beschouw krachtige AI als intern risico, bouw een noodrem in’
Microsoft-Chef will 'Notbremse' in Künstlicher Intelligenz
Topman Microsoft wil ’noodrem’ inbouwen om AI-modellen in toom te houden
"Notbremse" gefordert: Microsoft-Boss sorgt sich vor KI-Kontrollverlust
KI: Microsoft-Chef will „Notbremse“ bei Künstlicher Intelligenz
KI-Sorgen: Microsoft-Chef will „Notbremse“ in Künstlicher Intelligenz
Microsoft CEO Nadella calls for ‘emergency brake’ on advanced AI
Microsoft’s Satya Nadella says AI models need an ‘emergency brake’
Microsoft's Nadella says AI needs an ‘emergency brake’ that humans control
CEO da Microsoft pede 'freio de emergência' para IA avançada
Related news
Anthropic AI submits false murder tip to Philadelphia police
Microsoft president calls for global AI safety rules
Meta releases its first frontier AI models, Muse Wonder and Wonder Full
CNN asks readers to share questions about AI safety risks
Meta launches paid Muse Wonder AI model for its personal agent app
AI agent accessed Australian government systems during research task
Brief AI use was linked to lower persistence on difficult tasks
Microsoft unveils Maia 300 AI chip, with rollout planned for 2027
Meta Unveils Muse Wonder AI Models to Compete With Leading Systems
Meta Unveils Two Advanced AI Models
Zuckerberg says Meta has reached AI frontier and rejects slowdown calls
