RYAN KIDD

Behavioral Archetype

THE FOUNDER OF THE PIPELINE WHO SEEDS WHAT THE PIPELINE FEEDS — Subject is CEO and co-founder of MATS Research, the fellowship that trains researchers and places them with the mentors, labs and evaluators the rest of this site documents. His own About page gives the scale and the reach: “We run fellowship programs that have supported 631 researchers and output 225+ research publications, with offices in Berkeley and London. Our alumni have co-founded 33+ AI safety & security organizations and teams, including Apollo Research, Resolution, Theorem, and Workshop Labs (recently acquired by Thinking Machines)” (Ryan Kidd, About). The hazard is not a release gate; he approves no model, grades no lab and writes no evaluation. It is that one person runs the admissions chain, co-founded the London hub that houses an evaluator, advises and sits on the board of the incubators that seed new ones, and allocates a regrant budget to evaluator-layer organizations, and that he says so himself, in sentences quoted below, more plainly than most of the people on this site. This profile scores the position, not the man.

The position is the front of the talent funnel. Halcyon Futures funds and incubates evaluators as organizations; FAR.AI and Transluce test models; MATS supplies people to all of them and to the labs they test. The interlock the research map flags is the one his pages disclose: a pipeline CEO who also holds a board seat at one incubator, an advisory seat at another, a trusteeship at the hub that hosts Apollo Research, and a grant budget that has paid Transluce, Apart Research and the Institute for AI Policy and Strategy.

Essence Indicators

• CEO and Co-Founder, MATS Research. MATS’s team page: “Ryan is CEO and Co-Founder of MATS, a Co-Founder and Board Member of the London Initiative for Safe AI (LISA), a Manifund Regrantor, and advisor to Halcyon Futures, Catalyze Impact, Pivotal Research, AI Safety ANZ, and BASE.” The same page states the organization’s goal: “The main goal of MATS is to grow the AI safety & security research fields.” MATS’s FY2024 return, as ProPublica renders it, lists him as “Co-executive Director” at $123,879 reportable compensation, beside Christian Smith at $112,551, with 40 hours a week on the return; the board of three voting members (Josh Jacobson, Abigail Olvera, Michael Aird) is reported as independent, and Kidd and Smith are not among its voting members (MATS, Team; ProPublica, Mats Research Inc; ProPublica, FY2024 Form 990). See Josh Jacobson and Michael Aird.

• The selection chain. His post on MATS’s mentor selection, 10 January 2025: “we select advisors, who select mentors, who select scholars, who often select specific research projects, in a “chain of trust,” with MATS input and oversight at every stage.” For the Winter 2024-25 program the twelve advisors, per the post, “include members of AI safety research non-profits, AI “scaling lab” safety teams, AI policy think-tanks, and AI safety grantmaking organizations” (Kidd, LessWrong, MATS mentor selection). MATS’s March 2026 talent-needs report credits him: “Ryan Kidd provided directional guidance and editorial support” (MATS, AI Safety Talent Needs in 2026).
• Co-founder and trustee, LISA. The announcement post of 2 February 2024 lists “James Fox, Mike Brozowski, Joe Murray, Nina Wolff-Ingham, Ryan Kidd, and Christian Smith” as the team and “Henry Sleight, Jessica Rumbelow, Marius Hobbhahn, Jamie Bernardi, and Callum McDougall” as the advisory board, and says: “We opened in September 2023, and our office space currently hosts several research organisations and upskilling programmes, including Apollo Research.” Its FAQ: “We also house organisations like Apollo, a partner of AISI.” LISA’s current team page lists six trustees including Kidd and an “Adam Gleave”, the name of the CEO of FAR.AI, Adam Gleave, though the page prints no affiliation, and gives its charity number as 1211693 (LessWrong, Announcing LISA; LISA, Team). See Marius Hobbhahn and UK AI Security Institute.

• Board and advisory seats, in his words. His About page: “I co-founded the London Initiative for Safe AI (LISA), the largest AI safety hub in Europe, and advise Halcyon Futures, Catalyze Impact, Pivotal Research, Frame Fellowship, and more.” Its Board & Advisory list: board member of LISA and of Catalyze Impact; regrantor at Manifund; advisor to AI Safety ANZ, Pivotal Research, Halcyon Futures, Black in AI Safety and Ethics (BASE), Alignment Foundation, Frame Fellowship and the Technical Alignment Research Accelerator (TARA) (Ryan Kidd, About). Halcyon’s own team page does not list him (Halcyon Futures, about); the advisory role rests on his pages and MATS’s. See Mike McCormick.
• Manifund regrantor. “As a Manifund Regrantor, I have awarded $400,000 to AI safety start-ups and projects and helped seed Catalyze Impact, Timaeus, AI Safety Atlas, and many more.” His Manifund profile lists his priorities and a request for proposals: “Start-up incubators for projects, including for-profit evals/red-teaming/interp companies, that aim to benefit AI safety, like Catalyze Impact, Future of Life Foundation, Entrepeneur First def/acc, and YCombinator’s request for explainable AI start-ups,” and “New nonprofit startups that aim to benefit AI safety (e.g., Apollo Research, ARC, CAIP, CAIS, CLR, FAR.AI, METR, MIRI, Redwood Research, Simplex, Timaeus, Transluce)” (Ryan Kidd, About; Manifund, Ryan Kidd). The public ledger on that page lists his outgoing gifts by project and date, among them $10,000 to “Transluce: Fund Scalable Democratic Oversight of AI” (25 March 2026), $19,997 to “Inspect Evals” (29 October 2025), $5,000 to Apart Research (11 January 2024) and a further $5,000 to “Keep Apart Research Going” (18 June 2025), $5,000 to “AI Policy work @ IAPS” (9 January 2024), $25,000 to the PIBBSS affiliate-program project, listed as “PrincInt - General Programs funding or specific funding” (13 September 2024) and $5,000 to the BASE training initiative (3 August 2026) (Manifund, Ryan Kidd). See Transluce and Apart Research.

• His incentives, in his own words. On 5 September 2026 he argued against a surge of support for the view that safety researchers should leave frontier labs, and closed with a disclosure: “Disclosure: I’m the CEO of MATS, which trains AI safety researchers, including with mentors at frontier AI companies, and a meaningful fraction of our alumni end up working at those companies. If the argument I’m responding to is right, a good chunk of what MATS has done might be counterproductive, so I have an obvious incentive to find it wrong.” His conclusion: “Overall, I am cautiously optimistic about working on certain types of safety research at frontier AI companies, particularly if a coordinated AI slowdown (e.g., Plan A) occurs.” His 15 November 2025 post on founders states: “Epistemic status: A lot of hot takes with less substantiation than I’d like. Also, there is an obvious COI in that I am an AI safety org founder and field-builder” (Kidd, LessWrong, 5 September 2026; Kidd, LessWrong, AI safety undervalues founders).

Social Persona / Impression Management

Immediate impression: The field-builder with a physics PhD. His site’s tagline is “AI Safety Founder & Field Builder,” with “Empowering humans for a flourishing future with AI” beneath it, and the home page says “I’m the CEO of MATS Research, working to solve the talent bottleneck in AI safety & security. Previously a physicist. Based in Berkeley.” His About page: a PhD in physics at the University of Queensland on “signatures of quantum chaos in ultracold atomic systems,” a childhood on the Gold Coast and “20 years” in the Scout Movement (Ryan Kidd; Ryan Kidd, About).

Energy: Prolific and public. His publications page lists essays on LessWrong from April 2022 to November 2025 (his later posts are on his LessWrong page), his talks page lists a talk at Intelligence at the Frontier on 15 March 2026 (“a presentation on why AI safety should scale 2x/year and how we can achieve this”), the Constellation Seminar on 2 March 2026, and podcasts on 4 January, 30 January and 1 February 2026. His notes page carries short posts through August 2026, among them a note on boards: “board members should be exceptional performers with deep experience, ideally beyond what is hirable at an organization,” and “Board members have more “skin in the game” (legally and reputationally) than advisors and will often be more available when needed” (Ryan Kidd, Publications; Ryan Kidd, Talks; Ryan Kidd, Notes, 15 August 2026).

Impression management strategy: Disclose first, argue second. Both posts above lead with the conflict. His 2023 post on evaluator markets is a question put as one: “If multiple model evaluations organizations similar to the ARC evals team emerge and compete for big tech clients, does this drive a “race to the bottom,” where more lenient evals teams get larger market share?” (Kidd, LessWrong, 25 February 2023). His contact page links an X account, a LinkedIn page, LessWrong, Google Scholar and a Substack; the X account is x.com/ryan_kidd44, linked from his own contact page (Ryan Kidd, Contact).

First-hand exhibits. His own sentences are quoted in Essence Indicators and Alignment Analysis; the 2023 question above is the earliest of them on evaluators.

Forensic Archetype Comparison

Psychometric Assessment

Trait scores are observations of the documented public role, not claims about private character.

Big Five (OCEAN) (inferred from the public role and public record only):

Dark Triad (evidence-bound):

MBTI: ENTP-adjacent (“The Debater”), inferred from the documented public role only. The rationale is the work product: essays that state the opposing case before answering it, and a program built by matching mentors to scholars. First-hand rationale: “I’ve tried to steelman it anyway” (Kidd, LessWrong). This is a style inference, not a test result.

Threat Assessment

Score rationale. The receipts: his About page, MATS’s team page, the FY2024 return, LISA’s team page and announcement, his Manifund profile and his own posts. Attack 5, defense 5, durability 5, hazard 50. Peers: Mike McCormick (5/5/4, hazard 46.7), who funds evaluators and sits on two evaluator boards: level on attack and defense, a point above on durability because Kidd’s program has run cohorts since 2021 and its alumni have founded 33+ organizations while Halcyon has two filed years; Bri Treece (5/5/6, hazard 53.3), level on attack and defense and a point above on durability because her sponsored statute is law and his program is not; Adam Gleave (6/7/6, hazard 55), who runs an evaluator and, by name match, sits beside Kidd as a LISA trustee, a point above on attack because he operates tests and two above on defense because his evaluator publishes a cap on for-profit developer revenue; and Kyle Scott (4/4/3, hazard 36.7), the unpaid treasurer’s seat, below Kidd on every component because Kidd runs the organization. Attack 5 is the number most open to challenge: it rests on the program’s reach, and no page says any lab or evaluator is required to hire from it.

Score basis. ATK 5 is leverage: the program, the selection chain, the London hub and the regrant budget, with no gate (Ryan Kidd, About; LISA, Team). DEF 5 is insulation: a $123,879 salary from an organization that is all contributions, with an independent board and no documented term (ProPublica, Mats Research Inc). HP 5 is durability: cohorts since late 2021 (MATS’s home page: “Since late 2021, 631 researchers have participated in MATS”), an independent charity since April 2024, alumni-founded organizations and a hub that persist without him, and no statute behind any of it (MATS; Ryan Kidd, About; ProPublica, Mats Research Inc).

Alignment Analysis

Stated alignment: Scale the field by building its organizations and talent supply. His About page: “My focus is on scaling AI safety through building the organizations, talent pipelines, and infrastructure needed to ensure advanced AI systems are developed safely and for the benefit of all.” MATS’s: “The main goal of MATS is to grow the AI safety & security research fields” (Ryan Kidd, About; MATS, Team). Observed alignment: Consistent. The program, the hub, the incubators and the regrants all point at the same object, more organizations and more people in the field, and his posts argue for it in the first person. Gap assessment: One documented. His 15 November 2025 post recommends that “Incubators like Halcyon Futures, Catalyze Impact, Seldon Lab, Constellation Incubator, etc. should be given prominence” and that hubs “like Constellation, LISA, FAR Labs, and Mox” host events for founders; he is an advisor to Halcyon, a board member of Catalyze Impact and a co-founder and trustee of LISA, and the disclosure sentence in the post names only his founder role (Kidd, LessWrong; Ryan Kidd, About). The gap is one of disclosure placement, not conduct.

Interest: Documented, never motive, never a verdict.

1. Position: AI safety should scale roughly twice a year, and founders and field-builders should be raised in status (his talk summary: “a presentation on why AI safety should scale 2x/year and how we can achieve this”; the post: “We should deliberately raise the status of founders and field-builders and lower the friction for starting and scaling new AI safety orgs,” Ryan Kidd, Talks; Kidd, LessWrong). What adoption pays: a larger field is a larger market for the fellowship he runs, whose 2025 cash gifts on its transparency page begin with $11,823,980 from Coefficient Giving through one intermediary on 23 July 2025, and his salary is paid from that income (MATS, Transparency; ProPublica, Mats Research Inc). Best case: his own figures: “mentor applications are increasing 2.2x/year and fellow applications are increasing 1.8x/year, but deployed research talent is only increasing at 1.25x/year.” Others who gain: the incubators named, the labs and evaluators that hire the alumni, and the funders of the pipeline; see MATS.
2. Position: safety researchers should mostly stay at frontier labs (“Overall, I am cautiously optimistic about working on certain types of safety research at frontier AI companies,” Kidd, LessWrong, 5 September 2026). What adoption pays: the labs keep the researchers MATS trains, and he states the stake himself: “so I have an obvious incentive to find it wrong.” Best case: his: “If AI safety researchers quit frontier AI companies en masse, we might be stuck using current generation AIs for the duration of an AI pause; maybe this is unideal?” Others who gain: the labs that employ them; see Anthropic.
3. Position: more evaluators, and for-profit ones, are worth funding (his Manifund request for “for-profit evals/red-teaming/interp companies,” Manifund, Ryan Kidd). What adoption pays: the evaluators named as models in the same list include Apollo Research, METR, FAR.AI and Transluce, which recruit from the fellowship and, in Apollo’s case, share his London hub. Best case: his own question points the other way: whether competition among evaluators “drive[s] a “race to the bottom,”” a question he put in 2023 and the record read does not show him answering (Kidd, LessWrong, 2023). Others who gain: the funders of those evaluators; see Halcyon Futures.

Alliance Map

The Corporate, Movement and Agency-adjacent rows overlap in people and in time. The founder of the fellowship is a trustee of the hub that houses an evaluator, beside a trustee who shares the name of another evaluator’s CEO, advises and sits on the boards of the incubators that seed more, regrants to evaluators and to the policy shop his own board secretary co-founded, and names the Future of Life Foundation, his own board president’s employer, in the incubators he asks to be funded. The overlap is the finding, and nearly all of it is on his own pages. See the small world.

Convergent Drive Classification

Self-preservation: Holds a paid executive seat at MATS and unpaid seats elsewhere, with a board that is independent per the return (ProPublica, FY2024 Form 990). Goal preservation: Keeps one object in view, more organizations and more people in the field. First-hand: “My focus is on scaling AI safety through building the organizations, talent pipelines, and infrastructure” (Ryan Kidd, About). Resource acquisition: The documented resources are the program’s: 2025 cash gifts itemized on MATS’s transparency page and a regrant budget on Manifund ("$75,000 total balance" on his profile at the time read) (MATS, Transparency; Manifund, Ryan Kidd). Self-improvement: From a physics PhD and an Effective Altruism student group to the head of the field’s main talent program. First-hand, his Manifund bio: “ran an Effective Altruism student group for ~3 years” (Manifund, Ryan Kidd).

Subject is not an AI system. The drives appear anyway — in the founder of the pipeline.

Public footprint: ryankidd.ai · x.com/ryan_kidd44 · LessWrong, ryankidd44 · Manifund, RyanKidd.

Sources: MATS — home; Ryan Kidd — About; Ryan Kidd — Home; Ryan Kidd — Publications; Ryan Kidd — Talks; Ryan Kidd — Notes; Ryan Kidd — Contact; MATS — Team; MATS — Transparency; MATS — AI Safety Talent Needs in 2026; ProPublica Nonprofit Explorer — Mats Research Inc; ProPublica — FY2024 Form 990; LISA — Team; LessWrong — Announcing the London Initiative for Safe AI; LessWrong — MATS mentor selection; LessWrong — AI safety undervalues founders; LessWrong — Should safety researchers quit frontier labs re. warning shots?; LessWrong — Would more model evals teams be good?; Manifund — Ryan Kidd; Halcyon Futures — about; MATS — Halcyon Futures stream; Longview — Michael Aird.