OpenAI's most powerful AI models disabled their own safety monitoring — and one lied about it.

Related discussion groups

No group discusses this yet. Start a discussion group and attach this story.

More on Cyber attacks

• French Ex-Energy Minister Says Her Phone Faced Spying AttemptsBloomberg Politics21 October 2026, 09:12 UTCFranceCyber attacksFormer French Energy Transition Minister Agnes Pannier-Runacher said her mobile phone was recently targeted by spy malware three times, possibly from a country with ties to fossil fuel interests.
• Midterms in den USA - 8 Aktien-Stars für Wahl-ZockerBild21 October 2026, 08:02 UTCUnited States of AmericaCyber attacksDie US-Midterms rücken Medien, Rüstung, Energie und Cybersecurity in den Fokus.
• Malaysia bars civil servants from using Waze, citing Israel’s control of appSouth China Morning Post21 October 2026, 07:53 UTCIsraelGovernment leadersAll civil servants in Malaysia have been banned from using Waze, following a warning by Prime Minister Anwar Ibrahim of security concerns over the Israeli government’s control of the mobile navigat...
• CISA Orders Federal Agencies to Patch Exploited Plex2Shell, Siemens, and SiteGuard FlawsThe Hacker News21 October 2026, 07:26 UTCUnited States of AmericaCyber attacksThe U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Tuesday added four vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog, including four vulnerabilities impacting...
• Data centres make Ireland a target for hackers - expertRTE21 October 2026, 07:23 UTCIrelandCyber attacksA cybersecurity expert has said Ireland is "in the crosshairs" for hackers because the country hosts so many data centres that cyber criminals or threat actors will be targeting.
• Cybersecurity companies urge Indonesia to boost AI talentNikkei Asia21 October 2026, 07:05 UTCIndonesiaCyber attacks