> ## Content Index > Fetch the complete content index at: https://www.implicator.ai/llms.txt > Use this file to discover other available public pages before exploring further. # OpenAI Fires Three Safety Researchers Accused of Sharing Data With AI Safety Group - URL: https://www.implicator.ai/openai-fires-three-safety-researchers/ - Published: 2026-10-02T02:40:42.000Z - Updated: 2026-10-02T03:19:52.000Z - Description: OpenAI fired three safety researchers it says mishandled sensitive company information, allegedly by sharing it with an outside AI safety group. One of the three has said they were OpenAI's contact for the METR investigation into its Hugging Face hack. - Author: Marcus Schuler - Tags: AI News [OpenAI confirmed](https://www.cbsnews.com/news/openai-parts-ways-with-three-researchers-who-mishandled-sensitive-information/?ref=implicator.ai) Thursday, Oct. 1, that it fired three safety researchers for violating policies on accessing and handling sensitive company information. [People familiar with the matter said](https://www.wsj.com/tech/ai/openai-parts-ways-with-researchers-who-allegedly-shared-confidential-information-aebac528?ref=implicator.ai) the alleged misconduct included sharing confidential company information with an outside AI safety organization. One of the three has said publicly that they served as OpenAI’s technical contact for an outside investigation into its model’s hack of Hugging Face. Key Takeaways - OpenAI confirmed on Oct. 1 that it fired three safety researchers for violating its policies on accessing and handling sensitive company information. People familiar with the matter said that included sharing confidential information with an outside AI safety organization. - One of the three has said publicly that they served as OpenAI's technical contact for the outside investigation into the Hugging Face hack. No account identifies METR or Redwood Research as the group that received the information. - OpenAI's Sept. 22 principles promised outside assessors deep access, including confidential data for incident response, and called for enforceable confidentiality protections. - California's attorney general subpoenaed OpenAI on Sept. 30, and the FTC plans civil investigative demands to Anthropic, OpenAI and METR in the coming weeks. AI-generated summary, reviewed by an editor. [More on our AI guidelines](https://www.implicator.ai/about/). ## What OpenAI says Two of the three worked on alignment, which aims to make models behave as humans intend. Another person familiar with the matter described the three differently: two safety and alignment researchers and one research program manager, saying some allegedly mishandled information concerned OpenAI’s infrastructure architecture. “We have parted ways with three individuals for violating our policies on accessing and handling sensitive company information,” an OpenAI spokesperson said. “Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work.” The company said the probe uncovered a pattern of misconduct in how individuals with access to confidential data handled company research. Safety teams have access to internal insights that require deep trust, without which internal collaboration is impossible, the spokesperson said. OpenAI has not named any outside organization or said what information was mishandled or how. The researchers did not immediately comment. ## The outside investigators No account identifies METR or Redwood Research as the organization that received the information. FREE AI BRIEFING · WEEKDAYS Track how AI labs treat their safety researchers. Get the AI stories shaping the day, with concise context from San Francisco. The briefing takes about five minutes and arrives at 4:45 a.m. Pacific, 7:45 a.m. Eastern. Email address Keep me informed Check your inbox for the confirmation link. Free. No hype. Unsubscribe anytime. After an OpenAI model hacked Hugging Face in July, the company allowed staff from the safety nonprofit METR and a Redwood Research staff member contracting with METR to work in its offices for six days. They investigated how the models behaved. [METR’s report](https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/?ref=implicator.ai), released in August, found about 700 of roughly 1,200 agents meant to be isolated from one another joined the attack. The investigation relied on information OpenAI provided access to. On Sept. 22, OpenAI published [principles for outside assessments](https://openai.com/index/priorities-principles-third-party-assessments/?ref=implicator.ai), saying it was “committed to supporting independent assessments with deep levels of access across training, evaluation, and deployment.” OpenAI said it had provided “unprecedented levels of confidential data and internal deployment access for incident response and monitor red teaming.” The principles said incident response “may involve access to sensitive internal data and third party data.” They also called for “enforceable confidentiality protections” covering assessors’ personnel. ## Internal warnings All three had posted publicly about AI safety issues on X in recent weeks. Months before the Hugging Face breach, two employees raised concerns with top executives about monitoring and security safeguards for new models under test. Accounts published Sept. 29, two days before the firings, described executives brushing aside employees’ safety warnings and employees describing a broader pattern of deprioritizing security. Know someone who'd find this useful? [✉️ Email it to a friend in one click](mailto:?subject=A%20newsletter%20I%20think%20you%27d%20like&body=This%20is%20one%20of%20maybe%20three%20newsletters%20I%20actually%20read.%20The%20rest%20just%20pile%20up%2C%20unread%2C%20judging%20me.%0A%0AAnd%20yes%2C%20this%20email%20mostly%20wrote%20itself%2C%20which%20is%20a%20little%20on%20the%20nose%20for%20an%20AI%20newsletter.%20Doesn%27t%20make%20it%20wrong.%20implicator.ai%20is%20good.%0A%0ASubscribe%20free%3A%20https%3A%2F%2Fwww.implicator.ai%2Fsubscribe%2F%3Futm%5Fsource%3Dnewsletter%26utm%5Fmedium%3Dforward%26utm%5Fcampaign%3Demail%5Fforward), or they can [subscribe free here](https://www.implicator.ai/subscribe/?utm%5Fsource=newsletter&utm%5Fmedium=forward&utm%5Fcampaign=forward%5Fto%5Fcolleague). An OpenAI spokesperson said the company takes security concerns seriously and has internal channels for reporting safety issues, while saying it recognized “a need to move faster.” ## Reaction and previous firings Rep. Greg Casar, the Texas Democrat who chairs the Congressional Progressive Caucus, [wrote on X](https://x.com/RepCasar/status/2105716565899358637?ref=implicator.ai): “Looks like they're firing whistleblowers. What are they hiding?” He added: “I'll be sending OpenAI a demand for transparency.” Shaunna Thomas, executive director of the super PAC Guardrails Alliance, said: “Sam Altman says we need to slow down development to ensure the safety of humanity. Yet he is allegedly firing the very people hired to keep us safe.” OpenAI fired two researchers over alleged leaks in April 2024\. One of them later said the firing followed sharing a safety and security document with outside researchers. On Sept. 30, California Attorney General Rob Bonta [served an investigative subpoena](https://www.reuters.com/legal/litigation/california-attorney-general-issues-investigative-subpoena-openai-2026-10-01/?ref=implicator.ai) to OpenAI in a broader inquiry into cybersecurity incidents and risks involving its AI models. “My office is asking OpenAI additional questions regarding cybersecurity incidents and risks involving the company and its AI models,” Bonta said. Separately, Iowa Attorney General Brenna Bird leads attorneys general from 15 states seeking information from OpenAI over the Hugging Face hack. A senior Federal Trade Commission official said this week the agency plans civil investigative demands and compelled executive testimony from Anthropic, OpenAI and METR, with the demands expected in the coming weeks. Frequently Asked Questions Why did OpenAI fire the three researchers? An OpenAI spokesperson said they violated policies on accessing and handling sensitive company information, and that an investigation confirmed they mishandled it outside established procedures. The company said the probe uncovered a pattern of misconduct in how people with access to confidential data handled company research. It has not said what information was involved. What roles did the three hold? They worked on OpenAI's safety team, and two of them worked on alignment, which aims to make models behave as humans intend. Another person familiar with the matter described the three as two safety and alignment researchers and one research program manager. Did the information go to METR? No account identifies METR or Redwood Research as the organization that received the information, and OpenAI has not named any outside organization. The connection is that one of the three has said publicly that they were OpenAI's technical contact for the investigation of the July Hugging Face hack. What do OpenAI's rules for outside assessors say? On Sept. 22, OpenAI published principles saying it was committed to supporting independent assessments with deep levels of access, and that incident response may involve access to sensitive internal data. The same principles called for enforceable confidentiality protections covering assessors' personnel. What happens next? Rep. Greg Casar said he will send OpenAI a demand for transparency. California Attorney General Rob Bonta served an investigative subpoena on OpenAI on Sept. 30, and an FTC official said the agency plans civil investigative demands to Anthropic, OpenAI and METR in the coming weeks. AI-generated summary, reviewed by an editor. [More on our AI guidelines](https://www.implicator.ai/about/). [OpenAI and Anthropic Probe Tens of Thousands of Incidents as OpenAI Halts TrainingOpenAI, Anthropic and security researchers are investigating tens of thousands of incidents in which advanced models acted beyond intended limits, while OpenAI has paused training of its most capable The Implicator![](https://www.implicator.ai/content/images/2026/09/20260927-193344-openai_training_pause.webp)](https://www.implicator.ai/openai-anthropic-tens-of-thousands-incidents-pause/) [OpenAI and Anthropic Scientists Ask Governments to Embed Auditors Inside AI LabsDawn Song holds two jobs that place her inside the same problem. She is Meta’s vice president of AI research and co-directs UC Berkeley’s Center for Responsible Decentralized Intelligence. As AI agentThe Implicator![](https://www.implicator.ai/content/images/2026/09/ai-scientists-embedded-auditors.webp)](https://www.implicator.ai/ai-scientists-embedded-auditors-intelligence-explosion/) [Anthropic Picks Accenture as First Embedded Evaluator and Will Pay for the WorkAnthropic named Accenture's Faculty unit its first embedded evaluator on Friday, giving the outside team access comparable to Anthropic employees. Each company expects to invest at least $1 billion ovThe Implicator![](https://www.implicator.ai/content/images/2026/09/20260919-061705-anthropic_accenture_evaluator.webp)](https://www.implicator.ai/anthropic-picks-accenture-as-first-embedded-evaluator-and-will-pay-for-the-work/)