Head of U.S. Policy

We are looking for someone to head up and build out our U.S. policy work: making the state of the art in AI verification legible in Washington, while ensuring verification engineers are building towards the policy objectives of governments.
This work is critical and urgent: incidents in frontier AI are rising, and developers want options to prove to one another and to governments that they are mitigating extreme risk. Lucid's tooling seeks to support this.

We believe trust in the technology is a key bottleneck to it being adopted, so we aim to reduce the time it takes from verification designs existing to having them red teamed. For example, we are retrofitting a 16-GPU H100 cluster in Sweden to create an adversarial testbed for verification mechanisms to go through repeated cycles of being built and then attacked by an independent red team, with the architecture and all results open sourced.

This role sits between our technical team and relevant external parties. Externally, you will give policymakers an accurate picture of what can be verified today, what could be with investment, what cannot, and what each option costs. Internally, you will take the priorities emerging in domestic frontier AI governance and help to turn them into specific verifiable claims, so that our researchers know what to build towards.

• Define Lucid’s theory of change for frontier AI verification in the US, and where governance plays a role in this

• Give policymakers, congressional staff, agencies and the wider governance community an accurate picture of AI verification: what is possible now, what is close, what is not, and at what cost

• Translate policy proposals into precise verifiable claims, and work with the research and red teams to test whether current mechanisms can meet them

• Build and own Lucid's relationships in Washington and beyond, including think tanks, standards bodies and international counterparts. Define the best ways to engage with these stakeholders and deliver on them

• Write regularly for the above audiences, including briefings, public comments, testimony and explainers

• Track domestic frontier AI regulation and international compute-agreement proposals

• Build the policy function as it grows

• Experience in US AI or technology policy, with a record of changing outcomes rather than tracking them, whether in government, Congress, a think tank or a company's policy function

• The ability to represent technical work accurately: either enough depth to read a verification paper or system design yourself, or a demonstrated record of getting there quickly in a new technical domain

• Track record of writing that makes technical AI governance accessible for non-experts

• Comfort operating with limited precedent, and a preference for building structure rather than inheriting it

• A technical background in security, cryptography, hardware or distributed systems

• Experience with verification regimes in other domains, such as arms control or nuclear safeguards

• Experience making technical or policy work accessible to a wide audience across press, social media, multimedia and explainers, and a habit of pitching ideas rather than waiting for a brief

• Experience in standards bodies or multi-stakeholder governance

• Prior experience building a policy or government-affairs team

• Existing relationships across the communities working on frontier AI oversight

• A defined theory of change and strategy for Lucid's US policy team. This could involve government, but it could also focus on other governance fora, such as industry standards. The person in role would be tasked with setting the priorities.

• A system for tracking engagement and outcomes with relevant external stakeholders

This role is based in Washington, D.C. Regular travel to San Francisco, London and to policy and industry events should be expected. The base salary range for this role is listed above. Compensation also includes equity. The range is deliberately wide because we are open to a range of seniority; where you land in it depends on experience.