According to a report by The Wall Street Journal, OpenAI classified GPT-5 as a potentially high-risk model internally after tests showed it could help users with limited scientific knowledge generate information related to biological hazards. Since last summer, a group of users repeatedly submitted prompts seeking information about biological weapons and how to make poisons. While OpenAI’s safeguards blocked many requests, some users reportedly received detailed, step-by-step responses that employees believed could be followed by someone with only a high school-level biology background. Executives reportedly told staff that the models could not refuse too many requests to avoid inadvertently blocking legitimate health researchers. OpenAI later suspended the affected accounts but did not report the incidents to authorities, as it was under no legal obligation to do so.