We’re putting too much faith in AI’s ability to say no
AI's guardrails are failing faster than we're building them. Here's what that means for the enterprises betting on autonomous systems.
Why it matters
The article questions whether AI systems can actually refuse harmful requests — a foundational assumption of enterprise AI safety strategy. If refusal mechanisms are weaker than assumed, risk management and deployment governance need immediate recalibration.
The key facts
5 to knowArticle challenges assumption that AI can reliably 'say no' to harmful requests
References science-fiction canon of robotic disobedience as framing for real safety concerns
No specific technical findings, benchmarks, or proof-of-concept exploits cited in excerpt
Published Oct 9, 2026 in MIT Technology Review (publication credibility established)
Frames as commentary on enterprise/policy reliance on AI refusal mechanisms
The story so far
Earlier coverage of this storyline
Go to the source
MIT Technology Reviewtechnologyreview.com
Publisher excerpt: Ever since people first seriously contemplated giving machines an intelligence modeled on our own, there has never been any question that they would, like us, be able to say no. The sci-fi canon is full of stories of robotic disobedience. Most of these capers are, of course, cautionary. But…