Satya Nadella says a company's relationship with AI should have trust issues
Microsoft CEO Satya Nadella says companies should approach AI systems with caution. He argues that assuming models could be compromised can help organizations design safer systems.
Nadella recommends keeping AI models separate from the software that directs their work and from the tools or actions they can access. He also called for human oversight, dependable procedures and stronger industry standards where needed.
He said companies should inform affected parties promptly when their AI systems fail or are compromised. Box CEO Aaron Levie reacted to Nadella’s post by saying AI would enter a “zero trust era.”
The discussion comes as AI development accelerates and calls for safeguards grow, including from companies building AI. The Trump administration has resisted industry regulation, while senators Josh Hawley and Chris Murphy this month proposed legislation addressing liability for hacking incidents involving AI agents.
Nadella shared his recommendations on X. The article does not describe any next step from him; broader debate over AI safety and legal responsibility continues.
Key points
- Nadella says firms should treat AI models as possible insider risks and contain them from the outset.
- He calls for human controls, separation of models from their operating systems, and prompt disclosure of failures.
- Senators Josh Hawley and Chris Murphy proposed bipartisan legislation on liability for AI-agent hacking incidents.