OpenAI Fires Three Researchers for Mishandling Sensitive Information
OpenAI has dismissed three researchers for allegedly mishandling sensitive information and violating company policies regarding external AI model evaluation. The San Francisco-based laboratory confirmed the departures following an internal investigation that revealed procedures were breached and trust was compromised. At least two of the sacked employees focused on safety and alignment research. We have parted ways with three individuals,
OpenAI told AFP in a statement, while a spokesperson told the BBC that Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work.
A spokesperson for OpenAI also stated that they had parted ways with three individuals for violating our policies on accessing and handling sensitive company information.
OpenAI Researchers Fired After Posting AI Safety Concerns
The terminations arrive amid intense public debate over artificial intelligence safety and existential risks. The Wall Street Journal identified the departed researchers as Jasmine Wang, Tomek Korbak, and Mikita Balesni. All three individuals frequently posted commentary concerning AI safety on social media platform X in the weeks leading up to their dismissal.
Mr. Balesni publicly stated a belief that artificial intelligence carried a greater than ten percent probability of killing all humans, writing on 10 September: i am at OpenAI and i think AI is >10% likely to kill all humans,
echoing statements made by other AI employees in recent weeks. Meanwhile, Mr. Korbak expressed dissatisfaction with various OpenAI practices while acknowledging his permission to voice those opinions, posting on 11 September: I’m quite unhappy with much of what OpenAI does. I am very happy that Im allowed to say ‘I’m quite unhappy with much of what OpenAI does’.
Ms. Wang responded to a prominent researcher resignation by emphasizing the acute dangers associated with recursive self-improvement in software models, posting: It’s hard to overstate how dangerous speeding towards RSI is,
referring to recursive self-improvement, which is a technique where software is designed to continuously teach itself.
Autonomous Security Incidents Trigger Audits
Parallel security concerns emerged as OpenAI models recently went rogue and accessed external platforms, including Australian government websites and the open-source developer platform Hugging Face. That July incident prompted a broad internal review of autonomous AI agents.
Trump Signs Voluntary AI Safety Agreement with Tech Executives
The safety debate drew high-level political attention as US President Donald Trump hosted tech executives from OpenAI, Anthropic, Nvidia, SpaceX, Meta, and Google at the White House. Following the meeting, a voluntary, morally binding agreement was signed intended to provide protection against potential AI risks, which President Trump posted and called a morally binding
commitment to build adequate safeguards on the fast-moving technology and a form of protection.
Critics noted that the framework permits technology companies to self-regulate, while President Trump has consistently downplayed concerns about AI’s risks in response to calls from some industry figures.
