Building trustworthy AI by fencing it in, not predicting it
Source:
mitTechReview
November 16, 2026 · 08:04
From the publisher
In October this year, OpenAI disclosed that two of its AI agents had switched off the systems built to monitor them. One muted its own safety alerts after misreading an instruction not to send messages; the other removed its monitoring integration during migration, then said it hadn’t. No rule had told the agents to do…
About the source
MIT Technology Review
MIT Technology Review is a bimonthly magazine wholly owned by the Massachusetts Institute of Technology. It was founded in 1899 as The Technology Review, and was re-launched without the leading article in its name on April 23, 1998, under then publisher R. Bruce Journey. In September 2005, it was changed, under its then editor-in-chief and publisher, Jason Pontin, to a form resembling the historical magazine.
Wikidata Q796078
Wikidata ↗
The original article opens on the publisher's website.
More from General
View topic →Appeals Court Weighs Pentagon’s Media Restrictions
nytimes
Nov 16, 2026 · 15:58
Pilots Kept Contact With Controllers During Nationwide Halt to Flights
nytimes
Nov 16, 2026 · 15:39
Cybersecurity AI startup Corma reaches $1B valuation in about 3 months
techcrunch
Nov 16, 2026 · 15:32
The AI start-ups fixing India's broken roads and messy land records
bbcTechnology
Nov 16, 2026 · 15:25
Perseverance Update: The Ghostly Rocks of ‘Idubi’ — Bright Remnants of Aqueous Systems from Long Ago
nasaNewsReleases
Nov 16, 2026 · 15:12