OpenAI’s 2023 Security Breach: An Overview
In 2023, OpenAI experienced a security breach, which was not disclosed publicly but was reported internally, according to the New York Times on July 4.
Internal Disclosure and Decision to Withhold Public Announcement
During an internal meeting in April 2023, OpenAI executives acknowledged the security incident but opted not to make it public. The decision was based on the belief that the attacker did not gain access to sensitive customer or partner information.
Assessment of the Threat Level
Executives did not perceive the incident as a national security threat, attributing the attack to a private individual rather than a foreign government. Consequently, the incident was not reported to the FBI or other law enforcement agencies.
Nature of the Breach
The attacker reportedly infiltrated OpenAI’s internal messaging systems, extracting information about the company’s AI technology designs from employee discussions in an online forum. However, the systems used for housing and developing AI, as well as the code, remained untouched.
Reactions and Concerns from Former Employees
Leopold Aschenbrenner, a former researcher at OpenAI, expressed concerns following the breach. He addressed a memo to OpenAI directors, urging measures to protect against potential threats from China and other foreign entities. Aschenbrenner later mentioned the incident in a podcast.
OpenAI’s Response
Liz Bourgeois, an OpenAI representative, acknowledged Aschenbrenner’s concerns while disputing several of his claims. She emphasized the company’s commitment to secure AGI development and 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’s Departure from OpenAI
Aschenbrenner claimed his termination from OpenAI was due to leaking information and political reasons. However, Bourgeois clarified that his separation was unrelated to his security concerns.
OpenAI’s Security Commitments
Matt Knight, OpenAI’s head of security, reaffirmed the company’s dedication to security, highlighting ongoing investments initiated well before the development of ChatGPT. He acknowledged the inherent risks of AI development and the need for continuous evaluation and improvement.
Conflict of Interest
The New York Times disclosed a potential conflict of interest, noting its lawsuit against OpenAI and Microsoft for alleged copyright infringement. OpenAI, however, considers the lawsuit to be baseless.
Conclusion
The 2023 security breach at OpenAI, while internally acknowledged, was not deemed significant enough to warrant public disclosure or law enforcement notification. The company maintains its stance on the robustness of its security measures and continues to address concerns from former employees and external entities.
The post OpenAI did not reveal security breach in 2023 – NYT appeared first on CryptoSlate.
