Researchers “dismissed” from OpenAI: “We were expelled due to suspicious circumstances.” The company comments to Ofoq News

- 3 AI safety researchers at OpenAI said they were fired from the company, and questioned the reasons for their dismissal. The three researchers, Mikita Palisny, Tomac Korbak, and Jasmine Wang, worked on AI safety within the company and made sure the AI ​​systems did what they were meant to do in a safe way. But they say they were abruptly fired last week, weeks after they said they played a pivotal role in the investigation into an incident where OpenAI agents independently hacked AI company Hugging Face while it was undergoing testing.

Korbak said, via the “X” platform (formerly Twitter), that he was informed of his dismissal due to the way he communicated with the METR organization, which is concerned with the safety of artificial intelligence, with which OpenAI cooperated to investigate the Hugging Face incident. Korbak added that he was OpenAI's "key technical contact" with METR.

He added that he believes he was fired because "for months he had been raising safety concerns, as we lose the ability to monitor what AI systems are thinking, which is one of our best tools for detecting anomalies." He said he feared his firing would be used as an excuse for OpenAI to stop cooperating closely with METR, which is highly respected in the field of AI safety.

Palesny supported what Korbak said about the reasons for their expulsion, writing on the “X” platform that he was told that he “talks a lot with external safety organizations.” He said he understood OpenAI to insinuate that it had "leaked its intellectual property," which he denies.

These three expulsions follow a wave of serious security concerns expressed by those working in the artificial intelligence industry. While many former AI employees have publicly resigned over safety concerns, the three OpenAI researchers claim they were forced to resign for voicing similar concerns.

For her part, Wang said she was informed of her dismissal because she "saw an executive's email," which she said she obtained for recruitment purposes. Wang added that she repeatedly requested that her access be revoked, but when she accidentally saw a "sensitive email," she notified the CEO "within minutes" and notified OpenAI's IT team again.

“The reasons given to us for terminating our services are completely illogical,” Wang wrote on the X platform. She added: “We were not the first to be fired from OpenAI in questionable circumstances.

“Unless employees take a stand now against this type of maneuver, I fear we will not be the last to be fired.” The three researchers wrote to OpenAI management expressing their concerns and the reasons for their dismissal.

The Wall Street Journal was the first to publish news of the dismissal and the letter. “We do not believe that the trajectory of superintelligence can proceed safely if those closest to the risks can no longer work among themselves and with external parties with high confidence and high competence,” the researchers wrote.

OpenAI told Ofoq News Network that the researchers were fired not for “raising safety concerns, but for specific behavior that violated the company’s policies on handling sensitive information.” She added that "there were a number of violations that went beyond simple mishandling of information with an external evaluation group," she continued, and that "their dismissal was based on conduct that fell outside the scope of legally protected disclosures."

In a statement, OpenAI said it had terminated its relationship with the three researchers "for violating our policies regarding accessing and handling sensitive company information." A company spokesperson confirmed, “Our investigation demonstrated that these individuals mishandled sensitive information outside the scope of approved company procedures, which is a violation of our policies and a breach of the trust that is fundamental to our work.”

In a memo distributed to employees by OpenAI, one of the company's unnamed research leaders said they "strongly agree" with the three researchers on the importance of collaborating with third-party safety stakeholders. “I would like to make very clear that these decisions were not about raising safety concerns or expressing an opinion,” the official wrote, according to the memo.

“We don't fire employees just for raising concerns.” But the three fired researchers say they are already hearing from colleagues about the culture of fear developing within OpenAI.

“My former colleagues tell me they are confused about what to believe,” Palesny said. “They are afraid to speak up, afraid to search their personal phones for messages addressed to us or to third parties.”

“I fear that the prevailing fear of speaking and communicating with outside parties will lead to OpenAI being complacent on safety measures behind closed doors,” he added.