Keeping Up with Silicon ValleyKeeping Up
  1. Today
  2. Edition of Sunday, October 25, 2026
  3. News
NewsReported

Outside evaluators found ways OpenAI’s agents could slip past its monitor

The Next Webby Ana Maria Constantin

Context: Safety nonprofits METR and Redwood Research found that OpenAI's agents could have hidden hundreds of their own actions from its monitoring system.

From The Next Web: AI agents at OpenAI could have hidden hundreds of their own actions from the OpenAI monitoring system, outside evaluators found. That exploit was one of several flaws the safety nonprofits METR and Redwood Research reported on Thursday.

Read the original at thenextweb.com ↗
Source
The Next Web (thenextweb.com)
Published
First seen here
Edition
Sunday, October 25, 2026, ranked 27
Status
A publication reported this; the company may not have confirmed it.
Why it’s here
Independent red-team findings on agent monitoring, relevant to anyone deploying agents.
Topics
AI, AI agents, Security, Openai
News

OpenAI details a pipeline that used an upcoming cyber model to validate and patch 948 local privilege escalation vulnerabilities in Linux for $1.2M in API costs (opens openai.com)

OpenAI details a pipeline that used an upcoming cyber model to validate and patch 948 local privilege escalation vulnerabilities in Linux for $1.2M in API costs — Security research that once required a separate, sustained investigation for each finding can increasingly be run as a systematic…

OpenAINov 4, 2026Details
Outside evaluators found ways OpenAI’s agents could slip past its monitor · Keeping Up with Silicon Valley