Elizabeth Kelly

Founding director of the U.S. AI Safety Institute at NIST (2024–2025), now head of Beneficial Deployments at Anthropic, who argues that frontier models must be tested before deployment and that safety is the precondition for trustworthy AI adoption.

Roles

2021 – Feb 2024Special Assistant to the President for Economic Policy, White House National Economic CouncilHelped draft the domestic components of the 2023 AI executive order; start year inferred from the start of the Biden administration, not independently verified.Source →
Feb 2024 – Feb 2025Director, U.S. AI Safety Institute, National Institute of Standards and TechnologyInaugural director; signed pre-deployment testing agreements with Anthropic and OpenAI; left on 5 February 2025.Source →
Mar 2025 – presentHead of Beneficial Deployments, AnthropicLeads the team extending Anthropic's models to nonprofits, health and education organisations.Source →

On the record

25 Sept 2025 · Speech · United Nations, New York
By late 2026 or early 2027, we expect AI systems that will be smarter than Nobel Prize winners across most fields and capable of autonomously completing tasks that would otherwise take days or weeks. … But this brings both unprecedented opportunities and responsibilities that we must navigate together.
Head of Beneficial Deployments, Anthropic at the time
Speaking for Anthropic at a UN meeting on AI governance; published on the UN eStatements portal.
Remarks by Elizabeth Kelly, Head of Beneficial Deployments, Anthropic →
5 Sept 2024 · Interview · TIME
We're really building an enduring capacity here at the AI Safety Institute to make sure the state science of AI safety keeps up with the pace of AI development.
Director, U.S. AI Safety Institute, National Institute of Standards and Technology at the time
Elizabeth Kelly: The 100 Most Influential People in AI 2024 →
29 Aug 2024 · Statement · National Institute of Standards and Technology
Safety is essential to fueling breakthrough technological innovation. With these agreements in place, we look forward to beginning our technical collaborations with Anthropic and OpenAI to advance the science of AI safety.
Director, U.S. AI Safety Institute, National Institute of Standards and Technology at the time
Announcing the first government agreements for pre-release access to frontier models.
U.S. AI Safety Institute Signs Agreements Regarding AI Safety Research, Testing and Evaluation With Anthropic and OpenAI →
31 Jul 2024 · Interview · Center for Strategic and International Studies
We are really focused on testing – building out evaluation suites so that we can test new models prior to deployment. … There are so many open questions in AI safety right now – how these models work, why they produce some of the outputs they do.
Director, U.S. AI Safety Institute, National Institute of Standards and Technology at the time
Two separate remarks from the same conversation, joined with an ellipsis.
The U.S. Vision for AI Safety: A Conversation with Elizabeth Kelly, Director of the U.S. AI Safety Institute →
7 Feb 2024 · Statement · National Institute of Standards and Technology
The Safety Institute's ambitious mandate to develop guidelines, evaluate models, and pursue fundamental research will be vital to addressing the risks and seizing the opportunities of AI.
Director, U.S. AI Safety Institute, National Institute of Standards and Technology at the time
Statement on her appointment as the institute's first director.
U.S. Commerce Secretary Gina Raimondo Announces Key Executive Leadership at U.S. AI Safety Institute →