| Before | After | ||
|---|---|---|---|
| 12 | While we do not have definitive evidence that this model could meaningfully help a novice to create severe biological harm—our defined threshold (opens in a new window) for High capability—we have chosen to take a precautionary approach. 2025 System Cards Author OpenAI Keep reading View all Safety and alignment in an era of long-horizon models Safety Jul 20, 2026 Why teens deserve access to safe AI Safety Jul 16, 2026 GPT-Red: Unlocking Self-Improvement for Robustness Safety Jul 15, 2026 | 12 | While we do not have definitive evidence that this model could meaningfully help a novice to create severe biological harm—our defined threshold (opens in a new window) for High capability—we have chosen to take a precautionary approach. 2025 System Cards Author OpenAI Keep reading View all How enabling two settings tripled our scores on the ARC-AGI-3 benchmark Research Jul 29, 2026 Scientific computing in the age of agentic AI Publication Jul 28, 2026 Safety and alignment in an era of long-horizon models Safety Jul 20, 2026 |
Follow unlimited companies, monitor the clauses that matter across every platform, and get the full institutional analysis on what each change obligates you to do.