In brief
Three former researchers at OpenAI claim they were dismissed not for breaching protocol but for raising safety alarms. Their open letter, signed by Balesni, Korbak and Wang, points to a culture that punishes caution in pursuit of speed. OpenAI counters, citing an investigation that uncovered “significant breaches of trust”. The clash highlights an emerging rift in the tech sector between innovation pressure and ethical safeguards, prompting calls for independent safety assessors.
On 2 October a spokesperson for the AI company said an internal probe had found that three staff members mishandled sensitive information outside acceptable procedures. The investigators’ findings were delivered in a brief note that praised the researchers for their “trustworthiness” and denied any link to safety concerns.
The dismissed scientists, who had all worked in alignment and safety roles, refuse to accept the company’s narrative. In a tweet that felt more like a plea than a statement, former researcher Mikita Balesni said, ‘I believe we were fired for prioritising safety over the near‑term interests of OpenAI as a corporation.’ The claim was echoed in a joint open letter issued alongside a statement that the firings had stifled dialogue within the firm.
Balesni, Tomek Korbak and Jasmine Wang note that their departure has created a chilling effect; employees now risk speaking out on potential risks that could “leak into the organization’s future trajectory.” Korbak highlighted the loss of “the ability to monitor what AI agents think” as a key safety lapse that led to his exit.
OpenAI’s response remains concise: it stresses it was not about safety but about trust breaches, and it says contracts with external safety assessors are finalising. The company also points to plans to strengthen oversight in the coming weeks.
The dispute is unfolding at a time when the public and regulators increasingly expect AI firms to adopt ethical frameworks that keep pace with rapid development. Critics argue that unchecked growth raises silent threats – a concern amplified by incidents in other high‑profile projects. Whether the dialogue can return to a constructive footing will be a test for the sector’s future governance models.
