Helen Toner is an Australian AI policy researcher and the Executive Director of Georgetown University's Center for Security and Emerging Technology (CSET). CSET appointed her Interim Executive Director effective September 2, 2025, succeeding Dewey Murdick; before that she was the center's Director of Strategy and Foundational Research Grants (Source: cset.georgetown.edu; cset.georgetown.edu). She served on the board of the OpenAI non-profit from 2021 to November 2023 and was one of the four board members who voted to remove Sam Altman as chief executive on 17 November 2023, the event widely described as "the OpenAI board crisis." She has testified before several congressional committees and is a policy voice on AI oversight and US–China AI competition.
Background
Toner holds an M.A. in Security Studies from Georgetown University and a B.Sc. in Chemical Engineering and a Diploma in Languages from the University of Melbourne (Source: cset.georgetown.edu). She was an early hire at Open Philanthropy, where as a Senior Research Analyst she advised policymakers and grantmakers on AI policy and strategy. Before joining CSET she lived in Beijing, studying the Chinese AI ecosystem as a Research Affiliate of Oxford University's Center for the Governance of AI (Source: cset.georgetown.edu). She joined CSET at its 2019 founding; the center was launched with a founding grant from Open Philanthropy (Center for Security and Emerging Technology (CSET)). She joined the OpenAI non-profit board in 2021. TIME named her to its 2024 list of the 100 most influential people in AI (Source: cset.georgetown.edu).
Roles
CSET
As Director of Strategy and Foundational Research Grants, Toner helped define CSET's long-term priorities and led a multimillion-dollar technical grantmaking function; CSET describes her as "a driving force in shaping CSET's strategy and research agenda since its founding in 2019" (Source: cset.georgetown.edu). Her research areas there include compute, supply chains, AI capabilities forecasting, and governance. CSET's 2025 annual report records the late-2025 leadership transition in which she became executive director and Cara LaPointe joined as director of analysis (Source: cset.georgetown.edu). CSET describes itself as Washington's first think tank focused entirely on security and emerging technology and as the largest research organization in the world working on AI policy (Center for Security and Emerging Technology (CSET)).
OpenAI board (2021–2023)
On 17 November 2023 the OpenAI board fired Altman, stating he was "not consistently candid in his communications with the board." The four votes were cast by Ilya Sutskever, Tasha McCauley, Toner, and Adam D'Angelo; the Wall Street Journal reported that three of the four had effective-altruism affiliations, describing Toner as an executive of a center backed by a philanthropy dedicated to effective-altruism causes. When Altman returned days later, Sutskever, McCauley and Toner were replaced on the board, and D'Angelo was retained (How a Fervent Belief Split Silicon Valley—and Fueled the Blowup at OpenAI — Robert McMillan, Deepa Seetharaman (WSJ, November 22 2023)). During her board tenure she co-authored scholarly work on AI governance that Altman cited internally as a factor in tensions; she ultimately voted to remove him.
Toner has since given a public account of the episode. On The TED AI Show podcast in May 2024 she said the board had not been told in advance about the release of ChatGPT and learned of it on Twitter (Source: bloomberg.com). The complaint in Raine v. OpenAI quotes her as saying that Altman had been "withholding information," "misrepresenting things that were happening at the company," and "in some cases outright lying to the board." In a May 26, 2024 Economist essay with McCauley, "AI firms mustn't govern themselves," the two former directors wrote that "self-governance cannot reliably withstand the pressure of profit incentives" and that governments "must begin building effective regulatory frameworks now" (Source: economist.com). Her later public statements on the crisis also include appearances at TED AI.
In Musk v. Altman, a Toner deposition shown at trial on May 6, 2026 stated that Altman told the board three GPT variants had been submitted to the deployment safety board when only one had (Source: courthousenews.com). OpenAI's side used testimony from current and former insiders, Toner among them, to raise questions about Altman's credibility (Musk v. Altman (and OpenAI / Microsoft / Brockman)).
Congressional testimony
Toner has testified before several congressional committees (Source: cset.georgetown.edu). In June 2019 she testified before the U.S.-China Economic and Security Review Commission at a hearing on "Technology, Trade, and Military-Civil Fusion: China's Pursuit of Artificial Intelligence, New Materials, and New Energy" (Source: cset.georgetown.edu). On September 17, 2024 she testified before the Senate Judiciary Subcommittee hearing "Oversight of AI: Insiders' Perspectives" (Source: judiciary.senate.gov). There she said the biggest disconnect between AI insiders' perspectives and public perceptions concerned the idea of artificial general intelligence, and outlined six policy building blocks set out in her written testimony; her written testimony stated that her board experience had shown "how fragile internal guardrails are when money is on the line" (Source: techpolicy.press).
Other advisory roles
She was among the draft reviewers of the California governor's frontier-AI policy working-group report (The California Report on Frontier AI Policy).
Positions and statements
Toner emphasizes that the AI-risk problem is only partially technical, arguing that corporate governance, board independence, and institutional design are load-bearing elements of AI safety. She is critical of self-regulation at frontier labs, citing the OpenAI board episode as her primary empirical argument: internal governance at the leading lab failed visibly, which she has presented as evidence for external oversight (Source: economist.com).
On compute, she holds the CSET-aligned position that compute is the most tractable governance surface for frontier AI (see Compute Governance). On US–China competition, she supports export controls but resists hawkish framings, arguing for clear-eyed analysis of Chinese capabilities rather than racing rhetoric; she set out this view with Jenny Xiao and Jeffrey Ding in "The Illusion of China's AI Prowess" (Foreign Affairs, June 2, 2023), subtitled "Regulating AI Will Not Set America Back in the Technology Race" (Source: foreignaffairs.com). She advocates transparency and disclosure mandates, including pre-deployment testing, incident reporting, and whistleblower protections, many of which appeared in SB 53.
On her Rising Tide Substack she has argued that "the core challenge of AI alignment is 'steerability'" (April 3, 2025) and that "nonproliferation is the wrong approach to AI misuse" (April 6, 2025); both essays are cited in Law-Following AI. Her essay "'Long' timelines to advanced AI have gotten crazy short" is cited by Toby Ord for the observation that before ChatGPT "long timelines" meant no sign of AGI within 30 years, whereas many former proponents of long timelines now expect human-level AI within 10 or 20 years (Broad Timelines). In "The Term 'AGI' Is Almost Useless at This Point" (April 6, 2026), cited by the CNAS report Superpowers and AGI, she questioned the usefulness of the term (see Artificial General Intelligence (AGI)). Nita Farahany opened her AI law course with Toner's metaphor of AI governance as steering on a road that changes while one drives, with a fogged windshield and passengers who disagree about whether they are in a vehicle (Inside My AI Law & Policy Class 1: What is AI? (Farahany, August 2025)).
Selected works
- "The Illusion of China's AI Prowess," with Jenny Xiao and Jeffrey Ding (Foreign Affairs, June 2, 2023)
- Decoding Intentions: Artificial Intelligence and Costly Signals, with Andrew Imbrie and Owen Daniels (CSET, October 2023) (Source: cset.georgetown.edu)
- "How Developers Steer Language Model Outputs," with Thomas Woodside (CSET, March 8, 2024) (Law-Following AI: Designing AI Agents to Obey Human Laws (O'Keefe, Ramakrishnan, Tay & Winter, 2025))
- "AI firms mustn't govern themselves," with Tasha McCauley (The Economist, May 26, 2024)
- When AI Builds AI: Findings From a Workshop on Automation of AI R&D (CSET, January 2026) (Superpowers and AGI: How U.S.-China Competition Could Be Shaped by the Emergence of Artificial General Intelligence (CNAS, August 2026))
- Co-author of the 2026 shadow-evaluations study of frontier AI agents on unpublished research questions (Can AI agents conduct open-ended AI research? Early evidence from two case studies (Kirgis et al., July 2026))
- Co-translator, with Rogier Creemers and Graham Webster, of China's algorithmic recommendation provisions (Stanford DigiChina, January 2022) (China — Internet Information Service Algorithmic Recommendation Management Provisions)
- CSET research on AI governance, compute, and talent flows; congressional testimony across multiple hearings
Relationships
- leads: Center for Security and Emerging Technology (Executive Director)
- related: OpenAI — former board member; high-profile departure context; Sam Altman; Musk v. Altman (and OpenAI / Microsoft / Brockman) (deposition witness)
- supports: Compute Governance, Export Controls (AI), California SB 53, AI Safety Cases and Frameworks
- related: Lennart Heim — CSET-adjacent compute-governance collaborator
- related: Open Philanthropy (former employer), How a Fervent Belief Split Silicon Valley—and Fueled the Blowup at OpenAI — Robert McMillan, Deepa Seetharaman (WSJ, November 22 2023)