GPT-5.5 Biosecurity Vulnerability Assessment (openai.com)

• Targeted testing for biological hazard documentation
• Implementation of incentive-based adversarial testing
• Refining safety guardrails for frontier AI models

"A proactive security initiative targets GPT-5.5 to mitigate risks associated with biological threats. By launching a specialized bug bounty program, developers are identifying how generative AI models could be coerced into providing dangerous biological information. This effort is essential to ensure that advanced AI integration remains aligned with biosafety protocols, protecting critical societal infrastructure and scientific integrity against dual-use risks."