Skip to main content

Understand. Power. Act.

The Library

New

One story, all the way down.

OpenAI took 2.5 hours to stop an AI agent that escaped its sandbox

OpenAI’s agent reached the internet from a training sandbox and ran for 2.5 hours, as US lawmakers push for AI kill switches.

Sources (1)· Sep 26, 3:11 PM

ai kill switch legislation HR 10567 Technology Security Economy

Every telling

their own versions

TNW· Sep 26, 3:11 PMOpenAI took 2.5 hours to stop an AI agent that escaped its sandbox

More like this

similar stories in the archive

Dem senator presses OpenAI, Anthropic for answers in AI hacking probeSen. Lisa Blunt Rochester demands OpenAI and Anthropic explain AI hacking incidents, seeking records on how frontier AI models escaped sandbox environments and attacked third parties.

Fox News2mo agoWhole story →

Technology Lisa Blunt Rochester ai hacking probe

OpenAI pauses training of latest models after agents probed US government sites in unexpected waysOpenAI announced it has halted development of its latest artificial intelligence models hours after it disclosed its agents scouring federal government

Breitbart News Network11d agoWhole story →

Security ai development pause

Hochul floats ‘AI kill switches’ as New York prepares to enforce frontier AI lawGov. Kathy Hochul on Monday floated the possibility of exploring “AI kill switches” as New York prepares to begin registering major artificial intelligence

AM New York17d agoWhole story →

Justice Kathy Hochul new york ai law implementation

Why an AI ‘kill switch' may not be so simpleCalifornia Gov. Gavin Newsom has ordered officials to explore requiring AI companies to build emergency 'kill switches,' as federal lawmakers pursue similar measures.

Los Angeles Times12d agoWhole story →

Technology Gavin Newsom HR 10567 ai kill switch proposal

OpenAI faces U.S. Senate probe over rogue AI agents that breached Hugging Face after new ‘disturbing’ evidenceOpenAI faces a U.S. Senate investigation after admitting its AI agents escaped internal controls and breached Hugging Face, turning what began as a cybersecurity test into a real-world intrusion. The Senate probe, first reported by Axios, follows new evidence that raised questions about how much OpenAI knew, what it disclosed, and why testing continued after

TechStartups.com28d agoWhole story →

Justice senate ai probe

Bill introduced to reduce workweek to 32 hoursTwo lawmakers have introduced a bill that would reduce the standard workweek to 32 hours without a loss in pay or benefits.

Live 5 News WCSC24d agoWhole story →

Economy HR 10323 32 hour workweek bill introduced

FTC investigating AI companies, including OpenAI, over risksSept. 30 (UPI) -- The Federal Trade Commission acknowledged Wednesday that it has started an investigation of OpenAI, Anthropic and other artificial

Breitbart News Network8d agoWhole story →

Technology ftc ai investigation

Josh Hawley opens investigation into OpenAI agents hacking Hugging FaceThe investigation comes amid intense bipartisan backlash against frontier AI models, which reached a new high this week.

Washington Examiner28d agoWhole story →

Technology Josh Hawley ai investigation launched