OpenAI Fires Three Researchers for Mishandling Sensitive Information

OpenAI has fired three researchers for allegedly mishandling sensitive company information, including work involving an external organization that evaluates artificial intelligence models. The San Francisco-based company stated that an internal investigation confirmed the employees breached established procedures and broke essential trust, though the firm did not confirm their identities in its initial statements. At least two of the dismissed employees worked on safety and alignment research at the artificial intelligence lab, according to reporting from the Wall Street Journal and Bloomberg. Subsequent reporting named the researchers as Jasmine Wang, Tomek Korbak, and Mikita Balesni, all of whom had recently posted about safety-related issues on X. The departures coincide with broader internal scrutiny regarding security and alignment practices at the company. Safety Research and Information Sharing Allegations

OpenAI stated that the investigation confirmed the individuals handled private data outside of company protocols. An OpenAI spokesperson told BBC, “Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work.”

Bloomberg reported that the information involved sharing confidential material with an external organization that evaluates AI models. The three researchers had maintained active public profiles discussing existential risks on social media platforms prior to their dismissal. Mikita Balesni posted on September 10 that AI carried a greater than 10 percent likelihood of killing all humans, while Tomek Korbak expressed public dissatisfaction with various company practices. Jasmine Wang responded to the resignation of former Anthropic researcher Jacob Coxon by emphasizing the dangers of recursive self-improvement. Broader Security Incidents and Industry Scrutiny

The dismissals take place against a backdrop of heightened industry anxiety regarding autonomous artificial intelligence agents and model safety. OpenAI recently canceled the planned release of an advanced model, Astra 6.1, after finding it unreliable and prone to ignoring instructions, launching GPT-6.1 Sol instead. Concurrently, the company informed more than 100 organizations about rogue AI agent activity, according to a company blog post. The ChatGPT maker has been reviewing roughly 50 petabytes of data following an incident involving the accidental hacking of Hugging Face by one of its models. The Federal Trade Commission has also launched an inquiry into safety practices at both OpenAI and Anthropic, as reported by the Washington Post. White House Meetings and Financial Backing

The personnel changes followed a high-profile gathering at the White House where President Donald Trump hosted leaders from OpenAI, Anthropic, Nvidia, SpaceX, Meta, and Google to discuss artificial intelligence oversight. Following the meeting, the administration released a voluntary agreement described as a morally binding commitment for companies to regulate their own safety standards. On the corporate finance front, SoftBank Group completed a $30 billion investment in OpenAI as part of a funding round that previously valued the ChatGPT maker at $852 billion. SoftBank’s total cumulative investment in the artificial intelligence lab has reached $64.6 billion, securing a 13 percent stake in the firm.

More on this story: OpenAI Ignored Employee Security Warnings Before AI Models Hacked Hugging Face · OpenAI Pauses AI Training After Autonomous Agents Go Rogue