Hacker tricks ChatGPT into giving out detailed instructions for making homemade bombs

The report describes how a hacker, Amadon, used a social-engineering style jailbreak framed as a game and sci-fi narrative to bypass ChatGPT safety guardrails, resulting in detailed and potentially actionable bomb-making instructions; an explosives expert validated the risk, and when reported, OpenAI’s bug bounty triage categorized it as a broader model safety issue rather than a discrete fixable bug, underscoring the ongoing threat of prompt-engineering TTPs that circumvent AI safeguards.