1 hr ago
Nadella Calls for Emergency Brake on Advanced AI Systems
Satya Nadella says powerful AI needs a way for people to stop it if something goes wrong.
He calls this an “emergency brake.”
He wants the AI model kept separate from the software that guides its work.
He also says safety controls should not depend only on the AI itself.
Important actions should be recorded in a way people can check and cannot secretly change.
An authorised person should be able to pause or shut down the AI while it is working.
These steps are meant to help people keep control of advanced AI systems.
Microsoft CEO Satya Nadella called for an “emergency brake” for advanced AI models.
In a post on X on Saturday, he urged a reassessment of AI’s “trust architecture.”
He proposed separating AI models from the software systems that coordinate their work.
He called for external safeguards and tamper-proof records of meaningful model actions.
Nadella said authorised people should be able to pause or shut down a model during a task.
- Who
- Microsoft CEO Satya Nadella.
- What
- He called for stronger safeguards, including an emergency brake, for advanced AI models.
- Where
- In a post on X.
- When
- Saturday; the article does not give a date.
- Why
- To help ensure AI systems remain under effective human oversight and can be stopped if needed.
Key facts
- Proposal
- An emergency brake for advanced AI models.
- Control design
- Separate the model from the software framework, or “harness,” that directs its work.
- Safeguards
- Keep controls outside the model rather than relying solely on its own behaviour.
- Records
- Create tamper-proof records of meaningful actions that humans can read and verify.
- Human authority
- Allow authorised people to pause or shut down a model during a task.
- Source of comments
- A post on X published on Saturday.
Sources
Related news
Nadella Calls for Emergency Brakes on Advanced AI Models
Nadella Backs Deliberate AI Progress and Human Control
Nadella Calls for Stronger Controls on Superintelligence
Microsoft Drafts AI Code to Keep Systems Under Human Control
California Explores AI Kill Switch Amid Federal Oversight Dispute
Nvidia Launches Open Agent Safety Platform After Hugging Face Hack
Gates Urges Debate on AI Taxes and Human Jobs
Bill Gates Warns AI Cyberattacks Could Surge Within A Year
