GPT-5 bio bug bounty call

OpenAI has announced a bug bounty program specifically for GPT-5 related to biological data processing. This initiative aims to identify and fix vulnerabilities to ensure the model's safety and reliability. It matters because it helps improve the security and trustworthiness of advanced AI systems handling sensitive information.

ArchiveMajor

Signal trust

Single sourceEarly signal

PublishedFriday, September 5, 2025 at 10:45 AMSep 5, 10:45 AM

FreshnessArchive

Story ID#302

Back to feed Original report

Original article excerpt

Server-side extracted preview paragraphs from the original source.

OpenAI invites researchers to its Bio Bug Bounty. Test GPT-5’s safety with a universal jailbreak prompt and win up to $25,000.

As part of our ongoing efforts to strengthen our safeguards for advanced AI capabilities in biology, we’re introducing a Bio Bug Bounty for GPT‑5 and accepting applications. We’ve deployed GPT‑5 and are actively working to further strengthen safety protections for this and other models. We’re inviting researchers with experience in AI red teaming, security, or chemical and biological risk to try to find a universal jailbreak that can defeat our ten-level bio/chem challenge.

Opening the briefing

GPT-5 bio bug bounty call

Original article excerpt