NATE SOARES

Behavioral Archetype

THE MAN WHO WANTS THE OFF SWITCH — Subject is the president of the organization whose whole program is now one demand: stop the development of superintelligence, everywhere, by treaty. MIRI’s January 2024 strategy named the goal as an “international agreement to halt progress toward smarter-than-human AI” (MIRI, 2024 Mission and Strategy Update). In September 2025 he and Eliezer Yudkowsky published the book that put it in front of a mass audience, and in September 2026 MIRI endorsed a bill to write it into US law, on a statement “endorsed by Bourgon, Soares, and Yudkowsky on behalf of MIRI” (MIRI, Position on the Ban Artificial Superintelligence Act of 2026). The throughline is a technical researcher who became an executive and then a public advocate with a legislative ask. The hazard is the position that asks governments for the largest restriction on a general-purpose technology anyone has drafted, from an organization whose budget comes from a short list of named donors, while saying plainly that “We do not mean that as hyperbole.” This profile scores the seat, not the man.

Essence Indicators

• President and director, MIRI. MIRI’s team page lists him as “President” under Leadership beside Malo Bourgon (Chief Executive Officer) and Eliezer Yudkowsky (Co-Founder), and as “President” on its Board of Directors beside Blake Borgeson, Yudkowsky, Anna Salamon and Edwin Evans. His page: “Nate Soares is the President of the Machine Intelligence Research Institute, and plays a central role in setting MIRI’s vision and strategy.” The Form 990 titles on ProPublica’s page run Executive Director (tax years 2015 to 2020), Director and President (2021 and 2022) and President (2023 and 2024); Bourgon’s January 2024 post says that, “As we announced back in October, I have taken on the senior leadership role at MIRI as its CEO” (MIRI, Team; MIRI, Nate Soares; ProPublica Nonprofit Explorer; MIRI, 2024 Mission and Strategy Update). His own 31 May 2015 post: “Hello, I’m Nate Soares, and I’m pleased to be taking the reins at MIRI on Monday morning” (MIRI, Introductions). See Blake Borgeson.

• Before MIRI. MIRI’s bio: “Prior to MIRI, Soares worked as an engineer at Google and Microsoft, as a research associate at the National Institute of Standards and Technology, and as a contractor for the US Department of Defense.” The book’s media kit gives the shorter “previous experience at Microsoft and Google.” His 2015 post says he attended “my first MIRI workshop in December of 2013 while I was still working at Google” (MIRI, Nate Soares; If Anyone Builds It, Media Kit; MIRI, Introductions).

• MIRI’s money, from its filings. ProPublica’s extracted Form 990 data for the Machine Intelligence Research Institute Inc (EIN 58-2565917): tax year 2024, revenue $1,534,913, expenses $6,508,701, net assets $15,242,215; 2023, revenue $1,863,044, expenses $6,882,810, net assets $19,657,902; 2022, revenue $1,906,891, expenses $5,329,275, net assets $23,759,962; 2021, revenue $25,558,423, expenses $6,805,494; 2020, revenue $7,839,756, expenses $7,715,556. His reported compensation as the page prints it: $236,614 plus $10,635 other (President, 2024); $206,615 plus $9,959 (2023); $166,575 plus $9,976 (2022); $129,916 plus $9,246 (2021); $129,935 (Executive Director, 2020); $66,250 (2015). Beside him in 2024: Yudkowsky $599,970 (Chair, Head Researcher) and Bourgon $241,531 (Chief Executive Officer) (ProPublica Nonprofit Explorer). MIRI’s own December 2025 statement: it “expect[s] to hit December 31st having spent approximately $7.1M this year”; “Going into 2026, our budget projections have a median of $8M”; “Our goal at MIRI is to have at least two years’ worth of reserves on hand” (MIRI, 2025 Fundraiser).

• Who gives it money. MIRI’s Transparency page: “Our largest contributors, each of whom has given at least $500,000, are: An anonymous Ethereum cryptocurrency investor in 2018 and 2021. Open Philanthropy, a joint initiative with the philanthropic foundation Good Ventures. Vitalik Buterin, the inventor and co-founder of Ethereum,” then Jaan Tallinn, the Thiel Foundation, the Berkeley Existential Risk Initiative, Jed McCaleb, Loren Merritt, Edwin Evans, Richard Schwall and “A different anonymous Ethereum cryptocurrency investor in 2026.” Evans appears as “Edwin Evans, Cofounder of Linc Global” and, on the team page, as a director. MIRI’s 2025 fundraiser, “first in six years,” took in “just over $1.8M” with a $1.6M Survival and Flourishing Fund matching grant: “The first $1.6M raised will be matched 1:1 via an SFF grant” (MIRI, Transparency; MIRI, Team; MIRI, 2025 Fundraiser). See Open Philanthropy, Jaan Tallinn, Peter Thiel and Survival and Flourishing Fund.

• He is on the committee that recommends the money, too. The Survival and Flourishing Fund’s 2025 recommendations page lists its public Recommenders: “Malo Bourgon (Main Track) Nate Soares (Main Track) Nathan Labenz (Main Track) Seán Ó hÉigeartaigh (Main Track),” with others on the Fairness and Freedom tracks, in a round “for Funder Jaan Tallinn.” The same page’s grant table lists “Machine Intelligence Research Institute (MIRI) Main: $46,000 Freedom: $1,561,000 Fairness: $0 Mean: $0 $1,607,000” with a “Matching” figure of $1,392,000 (Survival and Flourishing Fund, SFF-2025).

• The book (16 September 2025). “If Anyone Builds It, Everyone Dies,” Little, Brown and Company, 272 pages, hardcover $30 (Hachette Book Group). Its thesis, as MIRI quotes it: “If any company or group, anywhere on the planet, builds an artificial superintelligence using anything remotely like current techniques, based on anything remotely like the present understanding of AI, then everyone, everywhere on Earth, will die. We do not mean that as hyperbole. We are not exaggerating for effect.” MIRI reports: “It was an instant New York Times bestseller, reaching #7 on the nonfiction list” (MIRI, 2025 Fundraiser). On the anniversary, MIRI wrote that the book “was held up in the chambers of Congress by Representative Brad Sherman in January,” that “We think the book aged well (unfortunately),” and that its Part Three answer is unchanged: “All over the Earth, it must become illegal for AI companies to charge ahead in developing artificial intelligence as they’ve been doing” (MIRI, One Year Closer, 16 September 2026). Palisade Research’s fundraiser: “We helped Nate Soares prepare for interviews following the publication of his recent book with Eliezer Yudkowsky” (Palisade Research).

• The policy asks, in documents with his name on them. (1) The January 2024 strategy: “international agreement to halt progress toward smarter-than-human AI, until humanity’s state of knowledge and justified confidence about its understanding of relevant phenomena has drastically changed.” (2) MIRI’s Problem page: “international community’s top immediate priority should be creating an ‘off switch’ for frontier AI development.” (3) The draft treaty on the book’s site, which says “We are not policymakers and we are not well-versed in international law,” then sets out an International Superintelligence Agency, a ban on “Any training run exceeding 1e24 FLOP or any post-training run exceeding 1e23 FLOP,” and a requirement that “all covered chip clusters (CCCs), as defined in Article II (i.e., a set of chips with capacity greater than 16 H100-equivalents)” sit in declared, inspectable facilities. (4) The September 2026 endorsement: “We hereby endorse the Ban Artificial Superintelligence Act of 2026 because it directly confronts the extinction threat that humanity is facing and would codify the primary policy goal we think the world needs: a ban on the development of superintelligence,” adding that “The bill lacks tracking and monitoring of AI chips” and that “Many MIRI staff want Earth’s civilization to build ASI eventually, after the point where ASI can be created in a fashion known in advance to be non-ruinous” (MIRI, 2024 strategy; MIRI, The Problem; If Anyone Builds It, Draft Treaty; MIRI, Ban ASI Act position). MIRI’s newsletter of 30 September 2026: “The ‘Bill to Ban Superintelligence’, introduced by Rep Casar and Senator Sanders, is shockingly good and includes steps toward an international ban. MIRI hereby endorses it” (MIRI, Newsletter #127).

• How the asks get to governments. MIRI’s fundraiser: “Thus far, the bulk of our engagement with policymakers and national security professionals has been done either by our CEO (Malo Bourgon), our President (Nate Soares), or the TGT researchers themselves”; “we have been told by various U.S. Members of Congress that the book had a valuable impact on their thinking. It remains to be seen how much this translates into action”; its Technical Governance Team “regularly provides input on RFCs and RFIs from various governmental bodies” and “in 2024 and 2025, TGT participated in the EU AI Act Code of Practice Working Groups”; and “Just four days ago, Malo was invited to provide testimony to a committee of the Canadian House of Commons.” Newsletter #127: “MIRI’s Technical Governance Team has been briefing members of Congress in the wake of the rogue AI events, consulting on legislation, recruiting policy advisors, and running a research fellowship program” (MIRI, 2025 Fundraiser; MIRI, Newsletter #127). See EU AI Act.

• Appearances. MIRI lists the interviews of “Nate and Eliezer” it favors: “Hank Green, The World Unpacked, Breaking Points, The Ezra Klein Show, and Making Sense with Sam Harris,” and says “Millions have now viewed interviews and appearances.” Its June 2026 newsletter: “Nate was one of six panelists in the Isaac Asimov debate hosted by Neil deGrasse Tyson, The Rise and Reckoning of AI”; “A clip from Nate’s Harvard Science Book Talk went viral on TikTok”; “Nate with Helen Toner at the 2026 CNBC CEO Council Summit”; long interviews with Novara Media and Americano. September 2026: “Nate did a two-hour interview on the Tucker Carlson show,” from which MIRI quotes him as saying “the bad news is that the bus is racing towards the cliff edge. The good news is that the bus driver is asleep”; and “Diary of a CEO hosted a discussion/debate with Ed Zitron, Roman Yampolskiy, Nate Soares and Andrew McAfee” (MIRI, 2025 Fundraiser; MIRI, Newsletter #126; MIRI, Newsletter #127). See Helen Toner.

• A dated exchange (The Argument, 14 September 2026). On the Sanders and Casar plan: “‘It looks like a pretty good start to me,’ he said during our Substack Live conversation last week”; on the economics of a pause: “‘Everyone who’s saying ‘what about the growth?’ has not actually grappled with the real situation,’ Soares told me.” The same piece sets MIRI’s plan, which “acknowledges ‘costs on industry and financial markets,’” against the AI Futures Project’s view that a slowdown is compatible with about 100 percent annual GDP growth in 2035, and the writer says: “I wonder if the quickest way to get policymakers on board is actually to appeal to the kinds of interests they have, not the principles they might eventually agree with” (The Argument).

• Technical record. Lead author of “Corrigibility” (AAAI-15 workshop, with Benja Fallenstein, Yudkowsky and Stuart Armstrong): “We call an AI system ‘corrigible’ if it cooperates with what its creators regard as a corrective intervention, despite default incentives for rational agents to resist attempts to shut them down or modify their preferences,” and “none have yet been demonstrated to satisfy all of our intuitive desiderata.” With Yudkowsky, “Functional Decision Theory: A New Theory of Instrumental Rationality” (arXiv 1710.05060): “This paper describes and motivates a new decision theory known as functional decision theory (FDT)” (MIRI, Corrigibility; arXiv).

Social Persona / Impression Management

Immediate impression: The earnest alarm-sounder with a spreadsheet. The record is a long, plain-spoken series of posts and interviews whose headline sentence is that everyone dies, delivered by an executive who also signs the budget, the policy briefing and the bill endorsement. His personal site and blog carry the older voice: a “communication handbook” and a blog of essays on motivation and conversation, the last on its first page dated 21 December 2019 (Minding our way; Minding our way, About).

Energy: High and public since September 2025: a book, a treaty draft, two-hour interviews, a debate panel, a Harvard talk, a CNBC summit, a live conversation with a newsletter, and a bill endorsement within three weeks of the bill’s announcement. MIRI’s comms team numbers “roughly seven full-time employees (if we include Nate and Eliezer)” (MIRI, 2025 Fundraiser).

Impression management strategy: Plainness as strategy. The anniversary post tells readers to “Say what you believe. Call out bullshit where you see it. Avoid the spin and doublespeak that are characteristic of AI companies and politicians” and says that “Many MIRI staff want Earth’s civilization to build ASI eventually,” which blunts the reading of the program as a rejection of the technology itself. The treaty draft opens by saying its authors “are not policymakers and we are not well-versed in international law,” and prints its own numbers as “best guess.” The exhibit that carries the most weight is a sentence about the opposition: the endorsement says the bill falls short because “The bill lacks tracking and monitoring of AI chips,” which positions MIRI as a supporter that asks for more than the sponsors do (MIRI, One Year Closer; MIRI, Ban ASI Act position).

• Social accounts and feeds. X @So8res, linked from his own site so8res.com, which his blog’s About page names as his “small website”; personal blog Minding our way with feed mindingourway.com/rss/, which says “This is a personal blog” and is not MIRI’s; MIRI’s X account is @MIRIBerkeley and its blog feed, which carries the posts under his byline, is intelligence.org/feed/, both linked from his MIRI page. No newsletter or Substack of his own found; his LessWrong history under the name So8res is not used here.

Forensic Archetype Comparison

Psychometric Assessment

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

Big Five (OCEAN):

Dark Triad (held low and evidence-bound; the score measures structural position, not character):

MBTI: ENTP (“The Debater”) — a theorist who now argues in public: a debate panel, a debate on a podcast, long interviews, and a treaty drafted as an argument with annotations and precedents. Rationale is the public record, not a typing session (MIRI, Newsletter #126; MIRI, Newsletter #127).

Threat Assessment

Score move: none (new profile). Calibrated against Stuart Russell (6/7/8, hazard 61, the one peer on identical values: a professor and textbook author with a research center and a long public argument about the same risk, but no organization to run and no recommender seat), Eliezer Yudkowsky (5/8/9, hazard 67, co-founder and co-author, with the founder’s insulation and twenty years more durability), Daniel Kokotajlo (6/6/7, hazard 63, runs a forecasting organization whose scenario work shaped the same debate), Jeffrey Ladish (7/6/7, hazard 58, a smaller organization with a published funding page and policy briefings) and Miles Brundage (6/8/7, hazard 68, runs an audit institute). Subject is one above Yudkowsky on leverage (the officer who carries the policy engagement and an SFF recommender seat), one below on insulation (the co-founder holds the larger compensation and the founder’s standing) and one below on durability (Yudkowsky’s twenty-year corpus); the hazard sits one point above Russell’s for the executive seat and five below Yudkowsky’s.

Score basis. Atk 6 is leverage: MIRI says the policy engagement has been done by “our CEO (Malo Bourgon), our President (Nate Soares), or the TGT researchers themselves,” the endorsement of the Ban Artificial Superintelligence Act of 2026 was “endorsed by Bourgon, Soares, and Yudkowsky on behalf of MIRI,” and he is a Main Track Recommender in the round that recommended $1,607,000 for MIRI (MIRI, 2025 Fundraiser; MIRI, Ban ASI Act position; SFF-2025); it is not a 7 because he holds no statutory, procurement or deployment seat and no government has adopted the treaty or the bill. Def 7 is insulation: MIRI grades no company, its board includes its co-founder and a listed donor, and the 2024 return shows $15.2 million of net assets (MIRI, Transparency; ProPublica); it is not an 8 because revenue swung from $25.6 million (2021) to $1.5 million (2024) and the donor list is short. Hp 8 is durability: the bestseller, the draft treaty, the 2015 and 2017 papers and an institution founded in 2000 (Hachette; draft treaty; arXiv); it is not a 9 because the treaty is described by its authors as illustrative and no state has taken it up.

Alliance Map

The overlap is the finding. The executive who asks governments for an international ban is at the same time a director of the organization the ban would benefit as a source of drafting and advice, a recommender in a fund that sends that organization money, and a co-author of the book that carries the message, with a donor on his own board; the donors, the fund and the policy aim all trace to a small set of named philanthropic and cryptocurrency sources. Each tie is printed on its own page; none is hidden.

Alignment Analysis

Stated alignment: “MIRI is a nonprofit with a goal of helping humanity make smart and sober decisions on the topic of smarter-than-human AI”; “The default outcome is lethal, but the situation is not hopeless; machine superintelligence doesn’t exist yet, and its creation can yet be prevented” (MIRI, 2025 Fundraiser).

Observed alignment: Writes the book, publishes the treaty draft, briefs legislators, endorses a bill and asks for more than the bill gives; MIRI’s own pages state each step. The one structural fact the pages print side by side is that the president sits on the committee that recommended MIRI a grant and on a board with a listed donor (SFF-2025; MIRI, Team).

Gap assessment: No documented gap between what he says and what he does. The positions are stated plainly, in the same words, across a book, a treaty draft, interviews and a bill endorsement. The structural item is the one the pages show: the organization that argues for the policy is funded by a short list of named donors, one of whom is on its board, and the president is a recommender in a fund that supports it. The record cannot settle whether the central claim is true, and for the author of a warning it does not have to: a claim is not made right or wrong by who funds the person who makes it.

Interest: (the faction’s stake in each position; a documented fact about the position, never a claim about motive.)

1. Position: an international treaty prohibiting the development of artificial superintelligence, with chip consolidation, inspection and a training ceiling (draft treaty).
   • What adoption would pay MIRI. No fee, contract or grant is attached to the draft. What the pages document is demand for its work product: MIRI’s Technical Governance Team is “briefing members of Congress … consulting on legislation, recruiting policy advisors,” and the fundraiser asks for “$6M” toward “a range of experiments” and growth in a budget projected at a median of “$8M” for 2026 (MIRI, Newsletter #127; MIRI, 2025 Fundraiser). The treaty creates an International Superintelligence Agency with a Technical Secretariat that does the kind of monitoring MIRI’s governance team studies.
   • His own best case, steelmanned. “We do not mean that as hyperbole. We are not exaggerating for effect. We think that is the most direct extrapolation from the knowledge, evidence, and institutional conduct around artificial intelligence today.” On cost: “Everyone who’s saying ‘what about the growth?’ has not actually grappled with the real situation” (MIRI, 2025 Fundraiser; The Argument).
   • Who else gains from the same rule. The compute and chip-verification researchers whose measurement work a threshold regime needs (GovAI’s Markus Anderljung co-wrote “To Govern AI, We Must Govern Compute,” and GovAI is, like MIRI, funded by the Survival and Flourishing Fund); the bill’s sponsors, Senator Sanders and Representative Casar; and the advocacy and communications groups that advise on delivery, among them Palisade Research (MIRI, Newsletter #127; GovAI, Annual Report 2025).
2. Position: a US-led global halt, and in the meantime an “off switch” (the Problem page: “the international community’s top immediate priority should be creating an ‘off switch’ for frontier AI development”; and MIRI’s newsletter that it published an op-ed “advocating for the US government to lead coordination of a global halt on frontier AI development”; MIRI, The Problem; MIRI, Newsletter #126).
   • What adoption would pay MIRI. An off switch needs “identifying the relevant parties, tracking the relevant hardware, and requiring that advanced AI work take place within a limited number of monitored and secured locations,” the Problem page says; MIRI’s governance team publishes the specifications it would build from.
   • His best case, steelmanned. The off switch “could also provide resilience to more limited AI mishaps,” so “we hope it will find broader near-term support than a full ban.”
   • Who else gains from the same rule. Any government seeking leverage over chip supply, and the evaluators and verification vendors whose products a monitored-compute regime would use; no page names a vendor.
3. Position: back the Ban Artificial Superintelligence Act of 2026, with a compute-threshold charter requirement and a pause until the agency is staffed (MIRI, Ban ASI Act position).
   • What adoption would pay MIRI. The bill “would codify the primary policy goal we think the world needs: a ban on the development of superintelligence”; the pages show no payment, only that the legislation’s goal is MIRI’s program.
   • His best case, steelmanned. “Banning artificial superintelligence (ASI), or variants of such a plan, is the only effective solution to avoid the ASI threat, at least in the near term,” because “Banning ASI requires intervening before ASI is created at all.”
   • Who else gains from the same rule. The agency the bill would create and its staff; the lawmakers who sponsor it; and, on the same compute thresholds, regulators who prefer a measurable trigger (“Training compute thresholds are often a good tool for AI governance because they are measurable ahead of time”).

Nothing above is a verdict that the positions are wrong. The finding is the alignment of interests: the person who asks for an international agency to monitor chips leads the organization that drafts the agency’s rules and briefs the legislators who would write them, funded by donors who also fund the other compute-governance groups.

Convergent Drive Classification

Self-preservation: MIRI’s stated aim to hold “at least two years’ worth of reserves,” a $16 million target against a projected $10 million at the end of 2025, and a board with its founder and a donor on it (MIRI, 2025 Fundraiser; MIRI, Team). Goal preservation: The goal, preventing superintelligence by agreement, is restated in the 2024 strategy, the book, the treaty draft, the bill endorsement and the anniversary post without change of wording (MIRI, 2024 strategy; MIRI, One Year Closer). Resource acquisition: The first fundraiser in six years with a $6 million target, a $1.6 million matching grant from the Fund on whose committee he sat, and a stretch target of $10 million (MIRI, 2025 Fundraiser; SFF-2025). Self-improvement: From research fellow in 2014 to executive director in 2015 to president and board member, and from decision-theory papers to a bestseller and a treaty draft (MIRI, Introductions; ProPublica; draft treaty).

Subject is not an AI system. The drives appear anyway — in the president of the institute that warns about them, which holds reserves, raises money, matches it and drafts the agency that would stop them in everyone else.

Public footprint: X @So8res (linked from so8res.com); personal blog Minding our way and feed mindingourway.com/rss/; MIRI blog feed intelligence.org/feed/; MIRI X @MIRIBerkeley. Papers: Corrigibility, Functional Decision Theory. Book site: ifanyonebuildsit.com.

Sources: MIRI, Nate Soares; MIRI, Team; MIRI, Transparency; MIRI, 2025 Fundraiser; MIRI, 2024 Mission and Strategy Update; MIRI, The Problem; MIRI, Position on the Ban Artificial Superintelligence Act of 2026; MIRI, If Anyone Builds It, Everyone Dies: One Year Closer; MIRI, Newsletter #126; MIRI, Newsletter #127; MIRI, Introductions (2015); MIRI, Corrigibility; If Anyone Builds It, Draft Treaty; If Anyone Builds It, Media Kit; Hachette Book Group, If Anyone Builds It, Everyone Dies; Survival and Flourishing Fund, SFF-2025; ProPublica Nonprofit Explorer, Machine Intelligence Research Institute Inc; Palisade Research, 2026 fundraiser; The Argument, The mechanics of an AI pause (with Nate Soares); Minding our way, About; arXiv 1710.05060, Functional Decision Theory; GovAI, Annual Report 2025.