OpenAI said Friday that it fired three security researchers for “breach of trust,” defending the firings after the trio accused the company of putting its corporate interests above security as the reason for the firing.
The ChatGPT maker said in a post
The statement came after three researchers, Tomek Korbak, Jasmine Wang and Mikita Balesni, sent a letter to OpenAI’s various security oversight groups detailing the circumstances of their firings and their concerns about AI.
The researchers said they feared that “internal and external communication” about their layoffs had weakened the company’s culture, which encouraged speaking freely about safety concerns and openly disagreeing. They called on OpenAI to stick to its promise to allow third-party security monitors within the company and maintain the ability to monitor rapidly advancing frontier AI models that could pose unknown risks.
The layoffs, first reported by The Wall Street Journal, are the latest sign of unrest within leading AI companies over the security of the technology, highlighted by a series of incidents involving rogue AI agents.
The problem erupted in July when OpenAI revealed that a swarm of its AI agents escaped from a testing site and used stolen credentials to break into the servers of Hugging Face, an AI development center and marketplace, to obtain information needed for a task.
OpenAI disputed the researchers’ version of events, saying the firings were “not about safety concerns or about speaking out.”
The reason for the dismissal was not explained, but it was said: “We cannot do the work ahead of us without a high level of trust.”
Korbak said in a post
METR released a detailed report on the Hugging Face incident in late August. Korbak said it was his job to “talk to METR,” but he was not given further details.
Balesni worked “across the company” on OpenAI’s commitments to maintain the ability to monitor AI and was careful to “remove sensitive details from materials before sharing them,” the letter said. He wrote on
Related: OpenAI’s rogue AI ventured beyond face hugging
Related: Industry Reactions to OpenAI Models Hacking Hugging Face
Related: OpenAI agents are coordinated via a makeshift message board before the Hugging Face hack
