Center for Security and Emerging Technology (CSET)

Theory of Change

CSET's theory of change: well-informed policymakers making better decisions about emerging technology reduces AI risk. From Open Philanthropy's founding grant writeup: "We think one of the key factors in whether AI is broadly beneficial for society is whether policymakers are well-informed and well-advised about the nature of AI's potential benefits, potential risks, and how these relate to potential policy actions."

The mechanism has two channels: (1) producing data-driven policy research that directly informs government decisions about AI, semiconductors, and national security, and (2) training people who move from CSET into government roles where they shape policy from the inside. Helen Toner (2019): "We have two primary goals -- produce research and recommendations on these topics and hopefully influence policy decisions, and train up the next generation of people who can think through these problems."

The founding team explicitly aspired to be "the RAND of AI" -- the institution that develops the conceptual frameworks policymakers use to think about AI-era security, analogous to how RAND shaped nuclear deterrence theory in the 1950s.

What They Do

CSET is a Georgetown University-based think tank focused on the intersection of national security and emerging technology. Since January 2019, it has produced 200+ publications, 190+ translations, and delivered hundreds of government briefings. Research topics include semiconductor supply chains, AI talent flows, export controls, Chinese military AI procurement, and frontier AI governance.

Core research products and policy influence:

• Semiconductor export controls: CSET researchers wrote extensively about using export controls to constrain China's AI chip access. Multiple alumni (Saif Khan, Ben Buchanan, Will Hunt, Emily Weinstein) then entered government positions at the NSC, BIS, and CHIPS Office where they directly shaped the October 2022 export controls and subsequent updates. The Jacobin described this as "taken right from the CSET-CNAS playbook."

• CHIPS and Science Act: CSET provided analysis informing the legislation. The 2022 year in review claims direct contribution.

• PLA AI procurement analysis: Recent work analyzing 2,800 Chinese military AI contracts, finding that private companies and universities receive the bulk of contracts rather than state-owned enterprises.

• Emerging Technology Observatory (ETO): A public data platform tracking AI research, patents, companies, and government programs. Funded in part by a $2M Google.org grant. Uses LLMs for entity resolution in large datasets.

• "Beyond Corporate Promises" (2025): Recent publication engaging directly with AI safety concerns (RSPs, evals, whistleblower protections) through a national security lens.

• Horizon Institute: Co-founded by CSET alumnus Remco Zwetsloot, this OP-funded ($3M+) nonprofit places tech fellows in congressional offices and government agencies. 12+ fellows working on AI policy across the Senate, House committees, DoD, DHS, and State Department.

Key People

Helen Toner, Interim Executive Director (Sept 2025). Former OP Senior Research Analyst. Former OpenAI board member -- voted to fire Sam Altman in Nov 2023, later revealed he had "lied to the board" and withheld information about ChatGPT's launch. TIME 100 Most Influential People in AI (2024). Positions herself as bridging the AI safety community and national security establishment: "I have my feet in two worlds."

Jason Matheny, Founding Director (2019-2021). Career arc: Future of Humanity Institute researcher -> IARPA Director -> CSET founder -> Biden White House (3 simultaneous roles in national security/tech) -> RAND Corporation CEO -> Anthropic Long-Term Benefit Trust member. Former director of research at Oxford's Future of Humanity Institute; published on existential risk. The single most connected individual in the OP-funded AI governance network.

Notable alumni pipeline: At least 6 former CSET staff have held senior government positions -- Saif Khan (NSC), Ben Buchanan (White House), Will Hunt (CHIPS Office), Emily Weinstein (BIS), Remco Zwetsloot (Horizon), Carrick Flynn (ran for Congress, now Brookings). Current team appears to be approximately 15 people, down from 50+ in 2022.

Money and Incentives

Total known funding: ~$112M+

Open Philanthropy represents approximately 94% of known funding. The OP grants were committed through 2025. No renewal announcements have been found. This is extreme single-funder dependency.

CSET operates under Georgetown University (EIN 53-0196603) and is not a separate legal entity. There is no independent financial reporting, making it impossible to verify annual spending, overhead, or reserves.

Structural relationships in the OP-AI network:

• Luke Muehlhauser: OP's grant investigator for CSET AND one of four Anthropic board members
• Holden Karnofsky: OP's former CEO AND married to Anthropic President Daniela Amodei
• Dustin Moskovitz: OP's primary funder AND early Anthropic investor
• Jason Matheny: CSET founder AND Anthropic Long-Term Benefit Trust member AND RAND CEO (where OP gave $15.5M)
• Joel Meyer: CSET advisory board member AND Anthropic's Head of Federal Civilian

CSET donated hundreds of thousands of dollars to CNAS, which also receives OP funding. OP funded Horizon Institute fellows placed at CSET, RAND, and other OP-funded think tanks.

Incentive analysis: CSET's business model is grants from a single funder whose broader portfolio includes significant financial and governance ties to frontier AI labs. CSET's research is substantive and often critical of AI companies, but the structural question remains whether an organization 94% funded by OP can fully scrutinize the network OP operates within.

What Others Say

Politico (Oct 2023): Detailed investigation of OP's AI influence network. CSET described as "funded almost entirely by Open Philanthropy." Deborah Raji (Berkeley): The existential risk focus is "almost like a caricature" and "going to lead to solutions or policies that are fundamentally inappropriate." Suresh Venkatasubramanian (Brown, co-author of White House AI Bill of Rights): Called it "speculative science fiction" and "fearmongering," compared to Eric Schmidt's influence operation. David Skaggs (former Office of Congressional Ethics chair): Horizon's structure suggests "an attempt to mask the program's ties."

Jacobin (Feb 2024): "Effective Altruists Are Stoking Tensions With China." Argues CSET alumni directly shaped Biden's semiconductor export controls, which amount to "a new cold war." Names the specific revolving door: Khan, Buchanan, Hunt, Weinstein. Claims CSET's policy papers were "taken right from the CSET-CNAS playbook" by the White House. The strongest substantive critique: not that CSET's research is bad, but that its policy direction (aggressive tech decoupling from China) may be counterproductive.

Zvi Mowshowitz (2024): Largely sympathetic analysis of Toner's OpenAI revelations. His critique of Toner is the opposite of Politico's -- that she understates existential risk concerns: "She takes this too far."

ChinaTalk (2022, 2025): Named CSET "Think Tank of the Year" in 2022. Jordan Schneider: "CSET has raised the bar for discourse in Washington."

Founders Pledge (2023): Recommends CSET as a donation target but notes "limited track record" and uncertainty about funding sustainability. Recommendation primarily based on OP endorsement and Matheny's credentials.

CSET's defense: Spokesperson Tessa Baker: "Open Philanthropy remains its largest donor" but CSET retains "complete and independent discretion over the research projects we conduct." Some CSET researchers "have emphasized a focus on AI's real-world harms in addition to long-term risks."

What's Absent

• Post-2025 funding: No announcements of OP renewal. The apparent staff reduction from 50+ to ~15 may reflect the funding cliff.
• AI safety community engagement: CSET has essentially zero presence on EA Forum, LessWrong, or Alignment Forum. $105M from a funder focused on "potential risks from advanced AI" went to an organization the technical AI safety community largely ignores.
• Transparent financials: No independent budget data because CSET is embedded in Georgetown.
• Non-Western critique: No Chinese academic or policy perspectives on CSET's export control advocacy found, despite enormous consequences for Chinese AI development.
• Impact evaluation: No independent assessment of CSET's policy influence claims beyond self-reported case studies.
• Biosecurity output: $3.33M specifically for a biosecurity research project (2021) with no visible published output.
• Staff reduction explanation: No public communication about why the team appears to have shrunk by ~70%.

Recommended Reading

1. Helen Toner on ChinaTalk (Dec 2025) -- The most candid current view of CSET. Toner discusses the jagged frontier thesis, dark arts vs good faith DC, her China trip, and hiring for frontier AI research. The interviewer is knowledgeable and pushes back productively. https://www.chinatalk.media/p/helen-toner-on-the-jagged-ai-frontier

2. Politico: "How a billionaire-backed network of AI advisers took over Washington" (Oct 2023) -- The strongest case against CSET's funding model and the OP influence network. Named critics, specific allegations, structural conflicts of interest. Essential counterargument. https://www.politico.com/news/2023/10/13/open-philanthropy-funding-ai-policy-00121362

3. Helen Toner on 80,000 Hours (July 2019) -- CSET's founding vision in enormous detail. Why immigration is the key AI policy issue, why the arms race framing is wrong, how CSET chose its focus. 26K words. https://80000hours.org/podcast/episodes/helen-toner-on-security-and-emerging-technology/

4. Jacobin: "Effective Altruists Are Stoking Tensions With China" (Feb 2024) -- The strongest substantive policy critique: CSET alumni drove export controls that may amount to a counterproductive tech cold war. https://jacobin.com/2024/02/effective-altruism-china-ai-us-policy

5. Open Philanthropy's founding grant writeup -- The funder's theory of change, identified risks, and explicit concerns about premature regulation. https://www.openphilanthropy.org/research/supporting-a-new-technology-security-policy-think-tank/