OpenAI has dismissed three researchers over what it described as the mishandling of “sensitive information” and breaches of company policy, including work connected to an outside organisation that assesses artificial intelligence models.
The San Francisco-based AI company did not publicly identify the employees, but the Wall Street Journal and Bloomberg reported that at least two had worked in safety and alignment, fields focused on making advanced AI systems behave reliably and remain under human control.
- Advertisement -
“We have parted ways with three individuals,” OpenAI told AFP in a statement.
“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 dismissals come as concern grows among AI researchers over the speed at which leading laboratories are developing increasingly powerful systems. Last month, Jacob Coxon, a 27-year-old researcher, resigned from Anthropic and warned that major AI companies — including OpenAI, where he had previously worked — were “gaming with our lives” by pushing ahead in the race.
The three researchers identified by the WSJ were Jasmine Wang, Tomek Korbak and Mikita Balesni.
In recent weeks, all three had posted frequently on X about questions surrounding AI safety and the potential risks posed by increasingly capable models.
“i am at OpenAI and i think AI is >10% likely to kill all humans,” Mr Balesni wrote on 10 September, echoing concerns expressed publicly by other AI employees in recent weeks.
As questions mounted over why staff at OpenAI and Anthropic were speaking so openly about their views, Mr Korbak joined the wider public debate.
“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’,” Mr Korbak wrote on 11 September.
Responding to Mr Coxon’s resignation, Ms Wang wrote: “It’s hard to overstate how dangerous speeding towards RSI is.” She was referring to recursive self-improvement, a technique in which software is designed to continually teach and improve itself.
The developments unfolded in the same week that leading US technology companies agreed to a voluntary safety pledge after meeting President Donald Trump at the White House.
Mr Trump described the agreement as a “morally binding” commitment to put adequate safeguards in place around the rapidly advancing technology.
Executives from Nvidia, Google, Meta, XAI, OpenAI and Anthropic signed the agreement.
