OpenAI names Dylan Scandinaro, who worked on AGI safety at Anthropic, as the company's “head of preparedness”, which has an up to $555K/year base salary
In December, OpenAI made headlines when it posted a job listing for a “head of preparedness” with an eye-watering base pay range of as much as $555,000 a year.
Context & Ripple Effects
OpenAI’s preparedness function grew out of its earlier team for testing models against catastrophic risks and an AGI-planning posture that treated advanced-model risks as potentially existential. The role has also seen organizational change: its prior leader was reportedly moved into research, making this appointment a concrete re-staffing of a specialized safety mandate.
The hire also draws talent from Anthropic’s AGI-safety work, while OpenAI already has governance mechanisms under which its board can withhold a model release despite management’s view that it is safe. That makes preparedness both an internal technical function and part of the lab’s external credibility case.
First-order effects
- Scandinaro takes ownership of OpenAI’s preparedness remit, restoring a named leader for assessment and risk-readiness work after the earlier reassignment of its prior head.
- The advertised pay ceiling makes the cost of recruiting senior frontier-model safety specialists visible, particularly for candidates with experience at a direct rival.
Second-order effects
- Anthropic and other frontier-model developers face stronger incentives to retain and recruit scarce safety researchers as preparedness leadership becomes a clearly compensated competitive role.
- OpenAI can more visibly connect model-risk testing to its release-governance process, increasing pressure for preparedness work to produce decision-ready evaluations rather than stand-alone research.
Third-order effects
- If labs continue to formalize and price senior preparedness roles this way, operational AI assurance may become a durable executive function alongside model research and deployment.
- The pattern could make safety capacity a more legible basis for strategic legitimacy, though the effectiveness of these roles will depend on their authority over deployment decisions.
The trend: Frontier AI labs are turning catastrophic-risk preparedness from a research initiative into a staffed, governable operating capability.