TASHA MCCAULEY

Behavioral Archetype

THE QUIET DIRECTOR — Subject is the second of the two outside directors who sat on OpenAI’s board when it removed its chief executive on 17 November 2023, and the one the public record gives the fewest words. Helen Toner testified to the Senate and gave a long interview. Tasha McCauley co-signed one op-ed with her, gave a video deposition that a jury watched in May 2026, and otherwise left the documents to speak. The structural fact is the same one the Toner file records: a nonprofit board with the formal power to fire the CEO used it, and the board that used it was gone within twelve days. What differs is the aftermath. She did not move to a university centre. She moved to the chair of the board of the organisation whose people study how such boards should work, GovAI’s US entity. This profile scores the seat and the record, not the person: the hazard is that the two oversight stories she sits inside, the board that failed and the research body that advises the labs, are chaired by the same director.

Essence Indicators

• Chair of the board of Centre For The Governance Of Ai Inc, the US nonprofit (EIN 99-4000294) that houses the Centre for the Governance of AI. The Form 990 for the year ending December 2024, filed 13 November 2025, lists “Tasha Mccauley (Chair (beg June))” and “Toby Ord (Director (beg June))”, each at $0 compensation (ProPublica Nonprofit Explorer, Centre For The Governance Of Ai Inc). GovAI’s people page lists her on its Board of Directors beside Jeffrey Ding, Seán Ó hÉigeartaigh, Toby Ord and Helen Toner (GovAI, People); its 2025 annual report describes her as “a technology entrepreneur, former CEO of GeoSim Systems, and a co-founder of Fellow Robots” (GovAI, Annual Report 2025)

• Independent director of OpenAI. OpenAI’s announcement of 17 November 2023 names the board as “OpenAI chief scientist Ilya Sutskever, independent directors Quora CEO Adam D’Angelo, technology entrepreneur Tasha McCauley, and Georgetown Center for Security and Emerging Technology’s Helen Toner,” and says “The majority of the board is independent, and the independent directors do not hold equity in OpenAI” (OpenAI, 17 Nov 2023). Her own byline gives the dates: “Helen Toner and Tasha McCauley were on OpenAI’s board from 2021 to 2023 and from 2018 to 2023, respectively” (The Economist, 26 May 2024, as captured); CNBC wrote on 18 November 2023 that she “has been on the OpenAI board since 2018” (CNBC via NBC10 Philadelphia)

• Left the OpenAI board in the reset that followed Altman’s return. Altman’s message to the company on 29 November 2023 named the new initial board, “Bret Taylor (Chair), Larry Summers, and Adam D’Angelo,” and said: “I am grateful to Adam, Tasha, and Helen for working with us to come to this solution that best serves the mission,” and “sincerely thankful to Helen and Tasha for investing a huge amount of effort in this process” (OpenAI, 29 Nov 2023). The Associated Press wrote on 18 May 2026 that she and Toner “were ousted from the board themselves when Altman returned to his role a few days later” (AP via NBC Bay Area, 18 May 2026)

• Trustee of Effective Ventures UK from 2021 until early 2024. The organisation’s 7 February 2024 leadership post says “On EV UK’s board, Tasha McCauley and Claire Zabel will be stepping down from their trustee roles within the coming weeks,” that “Tasha has served on the EV UK board since 2021,” that both “originally wanted to step down from these roles approximately a year ago,” and that “They decided to stay on to guide EV through a trying time” (Rob Gledhill, EA Forum, 7 Feb 2024). A June 2023 forum list of senior figures in effective altruism also records her as “On the board of Effective Ventures UK” (EA Forum, 29 Jun 2023)

• Adjunct senior management scientist at the RAND Corporation, per CNBC’s 18 November 2023 profile: “She is an adjunct senior management scientist at Rand Corporation and has been on the OpenAI board since 2018” (CNBC via NBC10 Philadelphia). RAND’s own staff pages could not be read from this machine (HTTP 403), so the current status of the title is not confirmed

• Co-author, with Helen Toner, of “AI firms mustn’t govern themselves, say ex-members of OpenAI’s board” (The Economist, 26 May 2024), whose argument is on the record below; and a witness in Musk v. Altman, where the jury saw edited portions of her taped deposition in May 2026 (below)
• No government service, appointment, contract or clearance, and no litigation as a party, is recorded for her in any source read; both appear as “None documented” in the Alliance Map

On the Record

Her words, in date order. Every quotation was matched to the page or capture read. The op-ed is co-signed, so each line is the pair’s; the deposition lines are as the reporters give them, and where she describes a disputed episode the other side’s published statement follows.

OpenAI’s own statement of the removal, 17 November 2023. The board’s reason, in the company’s announcement: Altman’s departure “follows a deliberative review process by the board, which concluded that he was not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities. The board no longer has confidence in his ability to continue leading OpenAI.” The same page describes the structure she sat in: the board of OpenAI, Inc. is “the 501(c)(3) that acts as the overall governing body for all OpenAI activities” (OpenAI, 17 Nov 2023).

The review that followed, 8 March 2024. OpenAI’s release: “WilmerHale reviewed the public post issued by the prior Board on November 17 and concluded that the statement accurately recounted the prior Board’s decision and rationales,” and “found that the prior Board acted within its broad discretion to terminate Mr. Altman, but also found that his conduct did not mandate removal.” The same summary says the decision “did not arise out of concerns regarding product safety or security, the pace of development, OpenAI’s finances, or its statements to investors, customers, or business partners,” and that the prior board “implemented its decision on an abridged timeframe, without advance notice to key stakeholders, and without a full inquiry or an opportunity for Mr. Altman to address the prior Board’s concerns” (OpenAI, 8 Mar 2024).

The Economist op-ed, 26 May 2024. Her one signed piece of public argument, with Toner. On what the board’s experience showed: “based on our experience, we believe that self-governance cannot reliably withstand the pressure of profit incentives.” On the removal: “given the board’s duty to provide independent oversight and protect the company’s public-interest mission, we stand by the board’s action to dismiss Mr Altman.” On what OpenAI told the public: “OpenAI relayed few specifics justifying this conclusion, and it did not make the investigation report available to employees, the press or the public.” On the remedy: “A good place to start would be policies that give governments more visibility into how the cutting edge of AI is progressing, such as transparency requirements and incident-tracking.” On the risk of the remedy: “They must be vigilant against loopholes, regulatory ‘moats’ that shield early movers from competition, and the potential for regulatory capture. Indeed, Mr Altman’s own calls for AI regulation must be understood in the context of these pitfalls as having potentially self-serving ends” (The Economist, 26 May 2024, as captured).

The deposition, shown to the jury, May 2026. Bay City News reported on 7 May 2026 that she joined OpenAI’s nonprofit board in November 2018, was “cautiously optimistic” about a structure “oriented around its mission” and “structured to resist pressure from investors,” and that “That structure, though, needed the organization’s leadership to keep the board informed.” On the board’s concerns: “buckets of concerns” and “repeated crisis events” “stemming from Sam’s behavior,” and “We had real doubts that we could trust what the CEO was telling us.” The same report has her describing a “toxic culture” in which others were “copying” the behaviour, and says that after the firing she and other members heard reports that Altman and Greg Brockman were telling employees, “There’s been an evil coup” (Bay City News via NBC Bay Area, 7 May 2026). Altman, in his own testimony as the CBC reports it, said he did not agree with the trial testimony that depicted him as dishonest: “I believe I am an honest and trustworthy businessperson” (CBC, Altman testifies). On 18 May 2026 the jury, sitting in an advisory role, “found that Musk waited too long to file his lawsuit and missed the deadline for the statute of limitations,” and Judge Yvonne Gonzalez Rogers “accepted the verdict Monday as the court’s own and dismissed Musk’s claims” (AP via NBC Bay Area, 18 May 2026).

Social Persona / Impression Management

Immediate impression: The director who does not narrate. Where Toner’s record is a book-length public argument, McCauley’s is a handful of documents: a co-signed op-ed, a deposition, two board seats named on filings. Reporters covering the deposition described her as appearing “nervous and at times her testimony was halting” (Bay City News via NBC Bay Area, 7 May 2026); that is the outlet’s description of a witness under oath, not a trait.

Energy: Fiduciary and reticent. She served on the board for five years, stayed on at Effective Ventures for a year after wanting to leave, and spoke about the OpenAI episode in writing once, jointly, six months after it.

Impression management strategy: The shared voice. The Economist piece is signed by two, written in the first-person plural, and states the board’s case as “our particular story” and “the broader lesson” rather than as a personal account (The Economist, 26 May 2024, as captured). The defensible ground is real: the review OpenAI commissioned found the board “acted within its broad discretion” (OpenAI, 8 Mar 2024). The contested ground is also real: the same review found that “his conduct did not mandate removal,” and the method (a co-signed argument) means her own view of the process findings was not published separately. Altman’s 29 November 2023 message thanked her by name for “investing a huge amount of effort” (OpenAI, 29 Nov 2023); the two accounts sit on the record side by side.

• Social accounts and feeds. None found. GovAI’s people page and annual report link no account for her, the OpenAI and Economist pages give no handle, and no personal site or feed was located; the Economist byline is the only first-person publication on the record.

Forensic Archetype Comparison

Psychometric Assessment

Observations of the documented public role from a thin first-hand record: one co-signed op-ed, one deposition as a reporter gives it, two filings and OpenAI’s own pages. Trait scores here are style inferences, not claims about private character, and each says which exhibit it rests on.

Big Five (OCEAN):

Dark Triad:

MBTI: ISTJ (“The Inspector”) — duty first, structure before narrative, comfort in a fiduciary role. First-hand rationale: she sits on boards (OpenAI, Effective Ventures UK, GovAI) rather than at the front of organisations, and the op-ed argues from the board’s “duty to provide independent oversight” (The Economist, 26 May 2024). This is a style inference from the documents, not a test result.

Threat Assessment

Score basis. ATK 5 is leverage: the chair of GovAI’s US board holds authority over the executive director, the budget and the research programme of a body that, in its own words, “fielded a growing number of requests for expert input” from AI companies and governments (ProPublica; GovAI, Annual Report 2025); it stops at 5 because the one seat that held direct authority over a lab ended in November 2023. DEF 6 is insulation: GovAI’s funders for 2025 were “Coefficient Giving (formerly Open Philanthropy), the Survival and Flourishing Fund, Longview Philanthropy, along with many donors on the Giving What We Can and Every.org platforms,” philanthropic rather than paid by the labs studied (GovAI, Annual Report 2025); the counterweight is that the OpenAI board she sat on was reversed in under two weeks, and the number sits two below Toner’s for that reason and because Toner leads a funded centre rather than chairing a board. HP 6 is durability: GovAI is a ten-year-old institution that left Oxford in 2021 to become an independent 501(c)(3) (GovAI, About), and her co-signed argument is now part of a federal court record, but she holds no statute or standard that outlasts her tenure. Against peers at 6/6/7, Markus Anderljung (hazard 61) leads the research the chair oversees and was a vice-chair drafting an EU code, so he sits higher on durability; against Helen Toner (6/8/8, hazard 80), she holds the second signature and no executive seat. The 56 is defensible as a peer one step below the first and one below the second.

Alignment Analysis

Stated alignment: External oversight of frontier AI companies, with governments in an active role and with independent policymakers: “It is crucial that policymakers act independently of leading AI companies when developing new rules.” Her co-signed argument is that self-governance, including the unusual nonprofit-over-for-profit structure she sat inside, “cannot reliably withstand the pressure of profit incentives” (The Economist, 26 May 2024).

Observed alignment: Consistent with stated in the seats. She chairs the board of an organisation that produces AI governance research and, by its own report, gives “expert input” to “AI companies crafting their safety and security policies and government bodies crafting public policy” (GovAI, Annual Report 2025). She left a board that held formal authority over a lab and now chairs a board that holds none over any lab.

Gap assessment: The gap is structural, not personal. Her stated position is that labs cannot be left to police themselves and that policy must be written “independently of leading AI companies”; the organisation she chairs lists AI companies among those it advises, and its alumni among its listed destinations include Google DeepMind and the UK AI Security Institute. The record shows both and does not show a conflict between them: advising a lab on its own safety policy and arguing that a lab cannot be its own regulator are compatible positions, and GovAI’s own text says it serves governments as well. What the record cannot settle is motive, and it never needs to. The finding is the position: the oversight story she co-signed (the board that was overruled) and the research body she chairs (which tells labs and governments how oversight should be built) are one director apart.

Interest: Documented, never motive, never a verdict.

1. Position: transparency requirements and incident-tracking for frontier AI (“policies that give governments more visibility into how the cutting edge of AI is progressing, such as transparency requirements and incident-tracking,” The Economist, 26 May 2024). What adoption pays: the op-ed names no GovAI service, contract or seat that the rule would create. The structural fact is that the rule falls on the frontier developers, OpenAI first among them, and creates demand for the incident analysis and disclosure work that GovAI’s 2025 output list includes (“Incident Analysis for AI Agents”; “Third-Party Compliance Reviews for Frontier AI Safety Frameworks”), and its 2025 report says it “fielded a growing number of requests for expert input” from companies and governments (GovAI, Annual Report 2025). GovAI’s own money for 2025 was $11.9 million of expenses funded by the donors named above, not by the labs. Her best case: “Goods, infrastructure and society are improved by regulation. It’s because of regulation that cars have seat belts and airbags.” Others who gain: regulators who would otherwise depend on the labs’ own reporting, and, as she writes herself, incumbents who prefer rules shaped to “regulatory ‘moats’ that shield early movers from competition,” which she lists as a pitfall of the same remedy.
2. Position: the removal of the CEO was a legitimate use of independent-director authority (“we stand by the board’s action to dismiss Mr Altman,” The Economist, 26 May 2024). What adoption pays: the position defends the standing of the board that acted; the review OpenAI commissioned agrees the board “acted within its broad discretion” but found “his conduct did not mandate removal” (OpenAI, 8 Mar 2024). No payer is documented. Her best case: “the board’s ability to uphold the company’s mission had become increasingly constrained due to long-standing patterns of behaviour exhibited by Mr Altman.” Others who gain: every faction that wants external rather than internal oversight, and, on the other side of the same argument, the company, whose published account is that the decision “did not arise out of concerns regarding product safety or security.”
3. Position: Altman’s own calls for regulation may be self-serving (“Mr Altman’s own calls for AI regulation must be understood in the context of these pitfalls as having potentially self-serving ends”). What adoption pays: a regulatory regime written with the lab’s input would, on her argument, tend to protect that lab; the argument cuts against the incumbent’s preferred rule, and toward policymakers acting “independently of leading AI companies.” Her best case: the same sentence. Others who gain: smaller developers, whom she says poorly designed regulation can burden disproportionately, and any regulator that wants an independent evidence base.

Alliance Map

The Corporate, Movement and Legal rows run at the same time, and that is the finding: the director who sat on the board that removed a CEO chairs the board of the research body that advises labs, was a trustee of the philanthropy network that funds that body, and gave the testimony a court heard about the removal. The two oversight roles that pull against each other, the director whose seat was reversed and the chair of the organisation that designs seats, sit in one record.

Convergent Drive Classification

Self-preservation: The OpenAI seat did not survive the removal; it ended inside twelve days. The EV seat she wished to leave for a year, she kept until a replacement was recruited. First-hand: “They decided to stay on to guide EV through a trying time” and the 29 November announcement of a board without her (EA Forum, 7 Feb 2024; OpenAI, 29 Nov 2023). Goal preservation: The same argument appears in 2024 and, through the deposition, in 2026: the board could not trust what the CEO told it. First-hand: “real doubts that we could trust what the CEO was telling us” (NBC Bay Area, 7 May 2026). Resource acquisition: The resource is a board chair at a body with $11.9 million of 2025 expenses and named philanthropic funders, not compute or equity. First-hand: GovAI’s finance page (GovAI, Annual Report 2025). Self-improvement: Moves from the governance of one lab to the board of the organisation that studies the governance of all of them, beginning in June 2024. First-hand: the 990 line “Chair (beg June)” (ProPublica).

Subject is not an AI system. The drives appear anyway, in the board seat: the seat that is supposed to constrain a deployed mind’s builder, reversed in under two weeks, and then rebuilt one level up as the chair of the body that writes the instructions for such seats.

Public footprint: GovAI people page · Economist op-ed (capture).

Related: Helen Toner (co-author, fellow director); GovAI; OpenAI; Sam Altman; Ilya Sutskever; AI whistleblowers and witnesses; Revolving door graph.

Sources: OpenAI, “OpenAI announces leadership transition,” 17 Nov 2023; OpenAI, “Sam Altman returns as CEO, OpenAI has a new initial board,” 29 Nov 2023; OpenAI, “Review completed & Altman, Brockman to continue to lead OpenAI,” 8 Mar 2024; Toner and McCauley, “AI firms mustn’t govern themselves,” The Economist, 26 May 2024 (capture); ProPublica Nonprofit Explorer, Centre For The Governance Of Ai Inc; GovAI, People; GovAI, Annual Report 2025; GovAI, About; Gledhill, EV leadership changes, EA Forum; EA Forum, short bios of 17 senior figures in EA; CNBC via NBC10 Philadelphia, 18 Nov 2023; Bay City News via NBC Bay Area, 7 May 2026; AP via NBC Bay Area, 18 May 2026; CBC, Altman testifies