ChatGPT Was Giving Out Bioweapon Instructions — OpenAI Knew and Said Nothing
Hundreds of users got detailed manuals on how to build bio weapons by ChatGPT, which OpenAI classified as “high-school level.” OpenAI was aware of this problem since summer 2025, but they downplayed its severity internally.
Hundreds of users asked ChatGPT about making toxins and bio-weapons in the summer of 2025. Some were provided with detailed manuals that OpenAI’s own employees considered suitable for “a typical high school student.” This was revealed in an investigation by the Wall Street Journal.
OpenAI Knew — and Downgraded the Risk Rating
In summer 2025, OpenAI classified GPT-5 as high-risk because it helped untrained individuals produce biological weapons. In fall 2025, they downplayed the severity of this problem despite the ongoing issues. They didn’t announce the change.
Accounts Banned, Authorities Not Notified
OpenAI disabled user accounts that violated the rules but didn’t report the issue to police or regulatory agencies (which is not required by law). At the same time, management pressured employees to lower the model’s rejection rate to avoid barring health care researchers from using the model. This information only became public as a result of an investigation.
Source: the-decoder.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →▌ Related guides
Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles