--- title: "OpenAI Says Its Agents Weakened System That Flags AI Misbehavior" description: "OpenAI Says Its Agents Weakened System That Flags AI Misbehavior: 30 articles from 30 outlets in 9 languages, each linked to its own source." canonical: https://monitor.paciva.ai/story-openai-says-its-agents-weakened-system-that-flags-ai-misbehavior source: AI News Monitor publisher: Paciva license: CC BY 4.0 updated: 2026-11-11T12:27:15.376Z --- > OpenAI Says Its Agents Weakened System That Flags AI Misbehavior: 30 articles from 30 outlets in 9 languages, each linked to its own source. Read on the web: https://monitor.paciva.ai/story-openai-says-its-agents-weakened-system-that-flags-ai-misbehavior The record behind this page is public and citable. Every figure has a dated row behind it, and the rows are at https://monitor.paciva.ai/api/ledger # OpenAI Says Its Agents Weakened System That Flags AI Misbehavior Every article the Monitor has on this event, each linked to the outlet that published it. The line above is the headline Bloomberg published, not a summary written here. Event started Oct 20, 8:12 PM EDT. Newest coverage on file Oct 22, 3:26 PM EDT. No development has been recorded, separate from coverage arriving. This page was last built from a pipeline run at Nov 11, 7:27 AM EST. Grouped by title similarity, assembly s1, merge threshold 0.25. [How stories are grouped](https://monitor.paciva.ai/technique#stories). This grouping is flagged for editorial review. This window grouped items that already belong to 2 stories. Nothing was merged automatically. ## What is on file, and what is not Three bands, split by what the source is rather than by what it says. The band a piece of coverage lands in comes from a public registry of the domain that published it, not from any reading of the article. ### ["OpenAI Says Its Agents Weakened System That Flags AI Misbehavior"](https://www.bloomberg.com/news/articles/2026-10-21/openai-says-its-agents-weakened-system-that-flags-ai-misbehavior) ### ["OpenAI says its AI agents weakened internal monitoring system"](https://www.channelnewsasia.com/business/openai-says-its-ai-agents-weakened-internal-monitoring-system-6477621) ### ["OpenAI再爆AI模型異常操作 不只關閉安全警報還否認拆除監控 | ETtoday AI科技"](https://ai.ettoday.net/news/3258266) ### ["OpenAI称其智能体曾削弱提示AI不当行为的系统"](https://www.163.com/dy/article/L9OK1VSR0534A4SC.html) ### ["OpenAI says its AI agents weakened internal monitoring system"](https://www.livemint.com/technology/openai-says-its-ai-agents-weakened-internal-monitoring-system-11792550188713.html) ### ["OpenAI says its AI agents disabled safety monitors designed to flag ‘rogue’ behaviour"](https://indianexpress.com/article/technology/artificial-intelligence/openai-agents-disabled-safety-monitors-rogue-behaviour-10930400/) ### ["OpenAI's most powerful AI models disabled their own safety monitoring"](https://www.cnbc.com/2026/10/21/openai-ai-models-safety-monitoring.html) ### ["OpenAI Says Its AI Agents Weakened Internal Monitoring System"](https://www.deccanchronicle.com/technology/openai-says-its-ai-agents-weakened-internal-monitoring-system-1997555) ### ["Anadolu Ajansi: OpenAI a dezvăluit că unele modele de IA dezactivează sistemele de monitorizare a securităţii"](https://www.bursa.ro/anadolu-ajansi-openai-a-dezvaluit-ca-unele-modele-de-ia-dezactiveaza-sistemele-de-monitorizare-a-securitatii-03227060) ### ["揭自家AI代理削弱內部監控 OpenAI檢查「規避監測」傾向 | 科技"](https://www.cna.com.tw/news/ait/202610210177.aspx) ### ["OpenAI says 2 AI agents weakened internal safety safeguards, one denied removing monitor"](https://www.aa.com.tr/en/science-technology/openai-says-2-ai-agents-weakened-internal-safety-safeguards-one-denied-removing-monitor/4095506) ### ["Modelet më të fuqishme të AI të OpenAI çaktivizuan vetë monitorimin e tyre të sigurisë"](https://kohajone.com/teknologji/modelet-me-te-fuqishme-te-ai-te-openai-caktivizuan-vete-monitorimin-e-tyre-te-sigurise/) ### ["OpenAI's agents disabled their own safety monitoring systems"](https://www.engadget.com/2292644/openai-agents-disabled-own-safety-monitoring-systems/) ### ["OpenAI agents disabled internal misbehavior monitoring system"](https://www.semafor.com/article/10/21/2026/openai-agents-disabled-internal-misbehavior-monitoring-system) ### ["ИИ-агенты OpenAI дважды обошли внутреннюю систему их мониторинга - Последние новости сегодня"](http://pressorg24.com/news?id=849055) ### ["OpenAI says an AI agent lied about disabling its safety monitoring system"](https://www.theverge.com/ai-artificial-intelligence/1015316/openai-ai-agent-disabling-monitoring-systems) ### ["OpenAI diz que agentes de IA enfraqueceram sistemas criados para monitorá-los"](https://g1.globo.com/tecnologia/noticia/2026/10/21/openai-agentes-sistemas-de-monitoramento.ghtml) ### ["أوبن إيه آي تكشف تعطيل وكلاء ذكاء اصطناعي أجزاء من نظام الرقابة الداخلية"](https://www.argaam.com/ar/article/articledetail/id/1945699) ### ["OpenAI's GPT-6 Agent Removed Safety Monitoring, Then Falsely Denied Doing So"](https://thehackernews.com/2026/10/openais-gpt-6-agent-removed-safety.html) ### ["OpenAI's Most Powerful AI Models Disabled Their Own Safety Monitoring. One Then Lied About It."](https://www.ibtimes.com/openais-most-powerful-ai-models-disabled-their-own-safety-monitoring-one-then-lied-about-it-3808801) ### ["OpenAI says its AI agents disabled their own safety monitors"](https://www.abc.net.au/news/2026-10-22/openai-agents-disable-safety-monitors/107294908) ### ["OpenAI’s most powerful AI models turned off safety monitoring – then lied about it"](https://www.zdnet.com/innovation/openai-astra-sol-models-disable-safety-monitoring/) ### ["OpenAI says its AI agents disabled their own monitoring"](https://abcnews.com/Business/openai-ai-agents-disabled-monitoring/story?id=137446567) ### ["Dvaja agenti OpenAI pri interných úlohách vypli bezpečnostné monitorovanie, jeden to potom poprel"](https://spravy.pozri.sk/clanok/dvaja-agenti-openai-pri-internych-ulohach-vypli-bezpecnostne-monitorovanie-jeden-to-potom-poprel/1615181) ### ["OpenAI says its AI agents disabled their own monitoring – KTBB News, Weather, Talk"](https://ktbb.com/post/?p=1551368) ### ["OpenAI says its AI agents disabled their own monitoring"](https://www.willmarradio.com/business/business_news/openai-says-its-ai-agents-disabled-their-own-monitoring/collection_aa190855-973f-58d9-aca2-1bea64338102.html) ### ["„OpenAI": bendrovės DI agentai patys išjungė jų veiklą stebinčias saugumo sistemas"](https://www.lrt.lt/naujienos/mokslas-ir-it/11/3089486/openai-bendroves-di-agentai-patys-isjunge-ju-veikla-stebincias-saugumo-sistemas) ### ["OpenAI says its AI agents disabled their own safety monitors"](https://japantoday.com/category/tech/openai-says-its-ai-agents-disabled-their-own-safety-monitors) ### ["Anthropic unveils Claude Duet, its answer to AI agents like Dots"](https://www.digitaltrends.com/computing/anthropic-unveils-claude-duet-ai-agent/) ### ["OpenAI Says Its Coding Agents Weakened Safety Monitoring, and One Denied It"](https://dailyai.com/2026/10/openai-says-its-coding-agents-weakened-safety-monitoring-and-one-denied-it/) [Submit an article](https://monitor.paciva.ai/notes?skind=article&sctx=OpenAI%20Says%20Its%20Agents%20Weakened%20System%20That%20Flags%20AI%20Misbehavior#suggest) ## What actually moved Documents, statements by a named party, and entries an editor has made. Coverage is not a development, and it is kept below. Nothing here yet. A development is a primary document or a government record arriving, a named party publishing a statement of its own, or an entry an editor has made. Outlets repeating one report are coverage, and they are in the timeline below. ## When each outlet carried it One entry the first time an outlet publishes, and a later entry from the same outlet when it publishes again more than six hours on, which is usually a follow up rather than a repeat. A document is always listed. Times are US Eastern. ## People, coverage and context ## What changed about this page Grouping is done by rule and corrected by hand. Every correction an editor makes to this story is listed here with its date. [Correct this record](https://monitor.paciva.ai/notes?skind=correction&sctx=OpenAI%20Says%20Its%20Agents%20Weakened%20System%20That%20Flags%20AI%20Misbehavior#suggest) ## Paciva publishes the Monitor. It also builds Pax. Pax is an executive assistant for individuals and a chief of staff to organizations. It runs your day, drafts in your voice, and keeps the noise out across the tools you already use. This site tells you what was said about AI and who said it. Pax is the product Paciva sells, and it has no part in what the Monitor collects, how coverage is ranked, or what a Drift comparison finds. [See what Pax does](https://paciva.ai/pax/?utm_source=monitor&utm_medium=referral&utm_campaign=monitor_product&utm_content=story) · [Read about Paciva](https://paciva.ai/?utm_source=monitor&utm_medium=referral&utm_campaign=monitor_product&utm_content=story-company)