A hacker said they purloined private details from countless OpenAI accounts-but researchers are doubtful, and the company is examining.
OpenAI states it's examining after a hacker claimed to have swiped login qualifications for 20 million of the AI company's user accounts-and put them up for sale on a dark web online forum.
The pseudonymous breacher published a puzzling message in Russian marketing "more than 20 million gain access to codes to OpenAI accounts," calling it "a goldmine" and providing potential buyers what they claimed was sample information containing email addresses and passwords. As reported by Gbhackers, the complete dataset was being offered for sale "for just a few dollars."
"I have more than 20 million gain access to codes for OpenAI accounts," emirking wrote Thursday, according to a translated screenshot. "If you're interested, reach out-this is a goldmine, and Jesus concurs."
If legitimate, this would be the third major security incident for the AI company considering that the release of ChatGPT to the general public. Last year, a hacker got access to the business's internal Slack messaging system. According to The New York Times, the hacker "took details about the design of the business's A.I. innovations."
Before that, in 2023 an even easier bug including jailbreaking prompts allowed hackers to obtain the private information of OpenAI's paying clients.
This time, however, security researchers aren't even sure a hack took place. Daily Dot press reporter Mikael Thalan wrote on X that he discovered invalid email addresses in the supposed sample data: "No proof (suggests) this alleged OpenAI breach is genuine. A minimum of 2 addresses were invalid. The user's only other post on the online forum is for a stealer log. Thread has actually because been deleted too."
No proof this supposed OpenAI breach is genuine.
Contacted every email address from the purported sample of login credentials.
A minimum of 2 addresses were void. The user's only other post on the online forum is for a stealer log. Thread has actually since been deleted as well. https://t.co/yKpmxKQhsP
- Mikael Thalen (@MikaelThalen) February 6, 2025
OpenAI takes it 'seriously'
In a declaration shared with Decrypt, an OpenAI spokesperson acknowledged the circumstance while maintaining that the company's systems appeared secure.
"We take these claims seriously," the spokesperson said, including: "We have actually not seen any proof that this is linked to a compromise of OpenAI systems to date."
The scope of the supposed breach stimulated issues due to OpenAI's massive user base. Countless users worldwide count on the company's tools like ChatGPT for business operations, academic purposes, and material generation. A legitimate breach might expose personal conversations, business tasks, and other sensitive information.
Until there's a final report, some preventive procedures are constantly advisable:
- Go to the "Configurations" tab, log out from all connected devices, and allow two-factor authentication or 2FA. This makes it essentially impossible for a hacker to gain access to the account, even if the login and passwords are compromised.
- If your it, setiathome.berkeley.edu then produce a virtual card number to handle OpenAI subscriptions. This way, it is simpler to identify and avoid fraud.
- Always keep an eye on the conversations kept in the chatbot's memory, and know any phishing attempts. OpenAI does not request for any personal details, and any payment update is always managed through the main OpenAI.com link.