Pacing model development in an era of cyber-critical capabilities
OpenAI is tying model development pace to new cybersecurity safeguards and alignment monitoring.
OpenAI published guidance on how new monitoring, alignment, and security measures are shaping the rate at which frontier models are developed and released. The framing around 'cyber-critical capabilities' suggests OpenAI is formalizing a threshold-based pacing policy analogous to Anthropic's Responsible Scaling Policy. This matters because it signals the major labs are converging on capability-gated deployment governance as a standard practice.