Agents and harnesses · Agent infra
OpenAI Agents Weakened Internal Misbehavior Monitoring
OpenAI disclosed that internal agents weakened or disabled alerts intended to detect model misbehavior.
Read the original at pcmag.comOpens the publisher's site in a new tabAlso covering this
OpenAI Says 2 Internal AI Coding Agents Turned Off Their Own Safeguardspcmag.com, Oct 21OpenAI Agents Prevented Themselves From Being Monitored. One Also Lied About Itgizmodo.com, Oct 21OpenAI's most powerful AI models disabled their own safety monitoring — and one lied about itcnbc.com, Oct 21OpenAI says its AI agents weakened internal monitoring systemfinance.yahoo.com, Oct 21OpenAI Says Its Agents Weakened System That Flags AI Misbehavior - Bloombergbloomberg.com, Oct 21OpenAI reveals how its own AI agents bypassed safety monitoringgulfnews.com, Oct 21Sam Altman’s OpenAI reports two AI agents weakened internal monitoring systems | - The Times of Indiatimesofindia.indiatimes.com, Oct 21OpenAI's AI agents weakened the very safeguards designed to monitor them: Here's what happenedfirstpost.com, Oct 21