OpenAI did not reveal security breach in 2023 – NYT

OpenAI had a security breach in 2023, which it did not publicly disclose, according to a New York Times report on July 4. The incident was shared internally in an April 2023 meeting, but executives did not reveal it to the public because the attacker did not access customer or partner information.

Executives also did not consider the incident a national security threat, as they believed the attacker was a private individual with no ties to a foreign government. Consequently, they did not report the breach to the FBI or other law enforcement agencies.

The attacker accessed OpenAI’s internal messaging systems and stole details about the company's AI technology designs from employee conversations. However, they did not access the systems where OpenAI develops its AI or the company's code.

The New York Times cited two individuals familiar with the matter as sources. They also mentioned former OpenAI researcher Leopold Aschenbrenner, who sent a memo to OpenAI directors after the breach, calling for measures to prevent foreign countries from stealing company secrets. Aschenbrenner alluded to the incident on a recent podcast.

OpenAI representative Liz Bourgeois acknowledged Aschenbrenner’s concerns but disputed his claims about the company's security and the incident. She stated, “We disagree with many of [Aschenbrenner’s claims] … This includes his characterizations of our security, notably this incident, which we addressed and shared with our board before he joined the company.”

Aschenbrenner claimed he was fired for leaking information and for political reasons, but Bourgeois said his concerns did not lead to his separation.

OpenAI's head of security, Matt Knight, emphasized the company’s long-standing commitment to security, stating they began investing in security measures years before ChatGPT. He admitted that AI development comes with risks that need to be addressed.

The New York Times noted a conflict of interest, as it is suing OpenAI and Microsoft over alleged copyright infringement of its content, a case OpenAI believes is without merit.