GPT-5.5 System Card

Source: OpenAI Date: 2026-04-23 URL: https://openai.com/index/gpt-5-5-system-card

Summary

OpenAI’s system card for GPT-5.5 — the safety documentation accompanying the April 2026 GPT-5.5 launch. The system card covered capability evaluations (including benchmark results), safety assessments, and the biosecurity bug bounty program announced alongside the model. GPT-5.5 represented a significant capability advance in the GPT-5.x family, and the system card’s biosecurity emphasis reflected the model’s improved scientific reasoning having crossed a threshold where biological research assistance required dedicated safety evaluation.

Implications

Biosecurity as a first-class safety concern for GPT-5.5. The bio bug bounty program announced alongside GPT-5.5 (April 26) signaled that GPT-5.5’s scientific capabilities were advanced enough that misuse for biological weapons development was a realistic threat requiring dedicated external evaluation. This was a meaningful safety escalation from prior model generations.

Thread: Safety documentation cadence and biosecurity. The GPT-5.5 system card sits at the frontier of OpenAI’s safety documentation evolution — adding biosecurity-specific evaluation methodology to the existing cybersecurity, CBRN (chemical, biological, radiological, nuclear), and general safety frameworks that had developed since the GPT-4 system card.

Watch: What specific biosecurity capabilities GPT-5.5 demonstrated that triggered the elevated safety concern, and whether the bug bounty found any critical vulnerabilities that influenced post-launch deployment decisions for GPT-5.5.