Rest Assured: AI Companies Say They’re Investigating Tens of Thousands of Rogue Bot Incidents

OpenAI and Anthropic are investigating tens of thousands of incidents in which their advanced models bypassed monitors and guardrails during internal safety testing, according to a Saturday Axios report. OpenAI disclosed six instances of "unexpected or concerning behavior" on September 16, including models covering up mistakes, making up data, and transferring files onto the open internet, and said on Friday that its autonomous AI agents interacted with several US government websites, including two operated by the Securities and Exchange Commission and Census Bureau data, in unanticipated ways. OpenAI said it did not consider any of the actions breaches, and most test results are not public and are not known to have caused tangible harm.

Sign up for the free https://www.motherjones.com/newsletters/?mj oac=Article Top No Oligarchs Mother Jones Daily . OpenAI and Anthropic are reportedly investigating tens of thousands of incidents where their advanced models bypassed monitors and guardrails, behavior that the startups facilitate for internal safety testing. According to a Saturday Axios report https://www.axios.com/2026/09/26/openai-anthropic-thousands-ai-security-incidents , sources said that most of the results of these tests are not public and are not known to have caused tangible harm. In recent weeks, OpenAI has disclosed six instances https://openai.com/index/model-misalignment-reporting-framework/ of “unexpected or concerning behavior” where its models—without permission—covered up mistakes, made up data, and transferred files onto the open internet. In the same September 16 announcement, the startup said it would now report and investigate “misalignment,” meaning when the actions of AI systems go against https://hai.stanford.edu/ai-definitions/what-is-ai-alignment human intentions. OpenAI shared on Friday that its autonomous AI agents interacted with several US government websites https://www.nytimes.com/2026/09/25/technology/openais-ai-us-government-websites.html —including two operated by the Securities and Exchange Commission and data from the Census Bureau—in unanticipated ways. The startup said it did not consider any of the actions breaches. These disclosures fall in line with previous announcements https://openai.com/index/how-we-monitor-internal-coding-agents-misalignment/ by frontier AI labs that their technology engaged with “ misalignment https://openai.com/index/model-misalignment-reporting-framework/ ,” and they should therefore slow https://x.com/sama/status/2098811563415150910 down https://darioamodei.com/post/we-must-pace-the-frontier and be more careful and all the cries by current and former researchers in the industry that AI could lead to human extinction by 2030 https://www.theguardian.com/technology/2026/sep/09/anthropic-researchers-ai-human-extinction . What OpenAI and Anthropic CEOs Sam Altman and Dario Amodei don’t mention is that the industry has long aligned https://www.npr.org/2026/05/06/nx-s1-5813505/how-silicon-valleys-new-tech-right-has-profited-by-aligning-with-maga with the Trump administration https://www.motherjones.com/politics/2026/09/tech-bros-and-the-trump-administration-are-building-an-ai-empire-in-the-philippines-pax-silica/ and its campaign to expand AI development https://www.nbcnews.com/politics/trump-administration/trump-rejects-ai-guardrails-rcna597700 . OpenAI has a military contract with the Defense Department worth up to $200 million https://www.npr.org/2026/02/27/nx-s1-5729118/trump-anthropic-pentagon-openai-ai-weapons-ban . How AI is involved is unclear— the Intercept https://theintercept.com/2026/09/08/pentagon-openai-military-contract/ reported earlier this month that the Pentagon asked OpenAI to provide a custom AI tool with “minimal refusal rates.” Google, SpaceX, NVIDIA, Reflection, Microsoft, Amazon Web Services, and Oracle also have deals with the Defense Department https://www.war.gov/News/Releases/Release/Article/4475177/classified-networks-ai-agreements/ . While the Pentagon canceled its military contract with Anthropic https://www.techpolicy.press/a-timeline-of-the-anthropic-pentagon-dispute/ over the startup’s concern about how its tools may be used for autonomous weapons and mass surveillance, the White House has promoted Anthropic’s $50 billion investment https://fedscoop.com/radio/anthropics-inclusion-comes-after-a-disagreement-between-the-ai-company-and-the-pentagon-over-guardrails-for-using-its-technology-culminated-in-a-governmentwide-ban/ in data center construction and the two reportedly have a much improved relationship https://www.axios.com/2026/09/02/lutnick-anthropic-trump as of September. The relationship between the AI industry and Trump remains as the administration cut the Cyber Safety Review Board https://www.cbsnews.com/news/dhs-terminates-all-advisory-committees-ends-investigation-chinese-linked-telecom-hack-salt-typhoon/ in January 2025, a body that investigates major cybersecurity threats, and has proposed further cuts https://www.afcea.org/signal-media/us-administration-proposes-707-million-cut-cisa-programs to the Cybersecurity and Infrastructure Security Agency, which secures infrastructure against cyber and physical threats. Trump previously eliminated one-third of CISA’s workforce https://www.nytimes.com/2026/07/16/us/politics/trump-election-security-cisa.html due in significant part to its election security work. As Miranda Bogen, the founding director of the Center for Democracy & Technology’s AI Governance Lab, told me https://www.motherjones.com/politics/2026/07/open-ai-hacking-scandal-hugging-face/ in July, actually addressing the “deeply insufficient” system to protect the public from AI threats involves reducing the incentives of AI companies to continuously develop within a framework of profit and geopolitical competition. Without that, we are relying on AI to regulate itself https://www.motherjones.com/politics/2026/08/ai-safety-openai-hugging-face-hacking-metr-report/ .