OpenAI staff warned of operational security lapses for years, sources say
openai
| Source: Mastodon | Original article
OpenAI staff have been flagging operational security shortcomings for years, with concerns raised internally before the launch of ChatGPT in late 2022, sources say.
OpenAI employees have been flagging operational‑security weaknesses for years, a Reuters investigation reveals. Five people with knowledge of internal discussions say staff raised concerns as early as the pre‑launch phase of ChatGPT in late 2022, warning that the company’s code‑base was overly open, security reviews were insufficient and insider‑threat safeguards were lacking.
The concerns were reportedly taken to senior executives and the board, but some employees left those meetings convinced that leadership was more intent on outpacing rivals such as Google and Anthropic than on tightening security. The allegations echo OpenAI’s own admission in October that a recording device had been found in an executive conference room, underscoring a pattern of lapses.
OpenAI responded on Monday with a blog post thanking the researchers who highlighted the issues and announcing a “next internal security sprint” aimed at shoring up defenses. The company said it had already taken steps to reduce the risk of similar attacks.
The story arrives on the heels of OpenAI’s recent public push on security, including the launch of a new cybersecurity model that it claims is better at preventing misuse of its tools. Observers will watch whether the announced sprint translates into concrete policy changes, tighter code‑review processes and clearer accountability structures. The next test will be how OpenAI balances rapid product rollout with the safeguards demanded by its own engineers and an increasingly security‑conscious market.
Sources
Back to AIPULSEN