FOUNDATION FOR AMERICAN INNOVATION

OLYMPUS opened an institutional file on the Foundation for American Innovation because it is the policy organization that appears at more than one table of the evaluator coalition at once. Its Vice President of Growth co-founded and sits on the board of AVERI, the auditing institute. Its President and CEO is listed as an affiliate member of Fathom, the sponsor of California’s evaluator-designation statute, and Fathom lists FAI as a partner. Its fellow Dean Ball co-authored AVERI’s flagship paper, took a fellowship at Fathom, drafted federal preemption text, and is now Head of Strategic Futures at OpenAI, which FAI’s own page names among its corporate supporters. This is not a psychometric profile. It is a supporters list, a filing, a board page and a set of drafts. The finding is the shape of the arrangement: a think tank funded by foundations and by the labs, whose scholars write the preemption and verifier proposals that those labs and the evaluators would live under, while FAI states that it “generally does not take positions as an institution.” Every tie is on a page the parties published. None is evidence that any funder shaped any draft.

Institutional Archetype

THE DRAFTING DESK WITH A SEAT AT EVERY TABLE — FAI describes its mission in one sentence: it “champions the technology, talent, and ideas essential to American prosperity, security, and flourishing” (FAI, About). It says it works “to make that possible”: “We help policymakers become more effective stewards of innovation, and we equip technologists and founders to better understand, engage with, and ultimately improve the machinery of governance.” Its outputs are “impact-driven research, sharp commentary, legislative engagement, and technical tools.” It runs a fellowship it names the Conservative AI Policy Fellowship and describes as “an eight-week, part-time fellowship for conservative policy professionals in Washington, DC” (FAI, Fellowships). It does not test models, grade labs or license anyone. What it produces for the evaluator coalition is text: bill drafts, comments, testimony and essays by its fellows, and people who move between its desks, the White House, the labs and the audit institutes. This profile scores the organization’s position, not any person.

Mandate & Origin

• Founding and legal form. FAI: “a 501(c)(3) nonprofit, established in Silicon Valley in 2014 as Lincoln Labs/Lincoln Network, and refounded in 2023 as FAI. Today we are a national technology think tank headquartered steps from the U.S. Capitol Complex” (FAI, About). ProPublica’s Nonprofit Explorer lists The Foundation For American Innovation, Washington, DC, “Tax exemption issued: March 2015” and “EIN: 47-2239840” (ProPublica Nonprofit Explorer).
• The advocacy affiliate. “FAI is a 501(c)(3) public charity focused on educational programs and research. For legislative advocacy and policy lobbying, we partner with our affiliated sister organization, FAI Action, a 501(c)(4) entity,” which is “a separate legal entity” (FAI, About). The first-quarter 2026 update: “in March we launched FAI Action, a new 501(c)(4) advancing energy and AI policy” (FAI, Q1 2026 update). A June 2026 open letter lists “Ian Mair, CEO, Foundation for American Innovation Action” (screendna.org).
• Leadership. FAI’s profile pages: Zach Graves “is president and CEO of the Foundation for American Innovation, as well as an ex officio member of the board. Previously he was head of policy at FAI,” and “a board member of the Institute for Progress” (FAI, Zach Graves). Max Bodach is Executive Vice President and Tim Hwang General Counsel (FAI, 18 June 2026; FAI, 29 June 2026). Grace Werner is “vice president of growth at the Foundation for American Innovation. Previously, she led strategy and U.S. policy research at AVERI” (FAI, Grace Werner). Samuel Hammond is listed as Director of AI Policy and Chief Economist (FAI, Q2 2026 update). Dean Ball is a non-resident senior fellow.
• Board, as the IRS return lists it. The FY2024 Form 990 as ProPublica summarizes it lists “Garrett Johnson (Chairman)” and, as board members, Kristen Soltis Anderson, Katie Biber, Michael Solana, Laurent Crenshaw, Allison Swope, Michael Kratsios, Aaron Ginn, Kmele Foster, Tamara Winter, and from April 2024 Meg Reiss, Reihan Salam and Mark Johnson, each at $0; the summary notes more people not shown. The FY2023 summary also lists Kratsios as a board member (ProPublica Nonprofit Explorer). Kratsios’s own questionnaire to the Senate lists the seat as 2023 to 2025; he has been Director of the White House Office of Science and Technology Policy since 2025, where Ball served (Michael Kratsios).
• The staff in the evaluator coalition. AVERI’s team page: “Grace Werner is a board member at AVERI and Vice President of Growth at the Foundation for American Innovation (FAI). Previously, Grace co-founded AVERI and served as Chief of Staff and lead U.S. policy” (AVERI, Team). AVERI’s flagship paper lists Dean W. Ball with the affiliation Foundation for American Innovation among its authors (AVERI, Frontier AI Auditing).

Funding & Backers

• What FAI says. “Our major philanthropic supporters have included Arnold Ventures, the Arthur N. Rupe Foundation, the Beth and Ravenel Curry Foundation, Coefficient Giving, the David and Lucile Packard Foundation, the Diana Davis Spencer Foundation, DonorsTrust, the Doris Duke Foundation, the Ethereum Foundation, the Filecoin Foundation for the Decentralized Web, the John S. and James L. Knight Foundation, the John Templeton Foundation, the Lynde and Harry Bradley Foundation, the Miles Foundation, the Proton Foundation, Recoding America Fund, Renaissance Philanthropy, the Sarah Scaife Foundation, the Searle Freedom Trust, the Silicon Valley Community Foundation, Stand Together Trust, Survival and Flourishing Fund, the Walton Family Foundation, the William and Flora Hewlett Foundation, and many others.” And: “Our corporate supporters and event sponsors have included a16z, Alphabet, Amazon, Anthropic, Astro Mechanica, Boom Supersonic, Cognition AI, DuckDuckGo, ExxonMobil, First Solar, Hadrian, Lightspeed Venture Partners, Meta, Microsoft, Mozilla, Narya Capital, New American Industrial Alliance, NRG Energy, Oklo, OpenAI, Paradigm Capital, Playground Global, Qualcomm, Rainmaker, Reflection AI, Shopify, Southern Company, T-Mobile, Uber, Ulysses, Variance, Waymo, XYZ Venture Capital, YCombinator, Yelp, Zanskar Geothermal, and various others” (FAI, About). No amount, date or per-donor share appears on the page. See Coefficient Giving, formerly Open Philanthropy, Survival and Flourishing Fund, Anthropic and OpenAI.
• What the filing shows. FY2024 revenue $5,149,927, of which “Contributions $5,106,107 99.1%”; expenses $3,719,664; net assets $3,427,251; the summary prints “Executive Compensation $1,248,309 33.6%.” FY2023 revenue was $3,679,704 with contributions of $3,542,134, 96.3 percent. The FY2024 return was filed on 12 November 2025; no FY2025 return was available (ProPublica Nonprofit Explorer). FAI’s own account of 2025: “We significantly grew our staffing, revenue, and standing” (FAI, Q1 2026 update).
• The gift policies. “FAI does not accept donations from foreign governments,” and it has “adopted a strict policy of refusing all gifts from entities with substantial ties to any country listed as a U.S. foreign adversary under 15 CFR 791.4 (China, Cuba, Iran, North Korea, Russia, Venezuela)” (FAI, About).
• The open voice. “The Foundation for American Innovation does not adhere to a one-voice policy. FAI staff are free to have their own views, endorse positions, talk to media, and openly disagree with their peers as long as this is consistent with FAI’s work, values, and mission. As a result, FAI generally does not take positions as an institution, instead relying on the expertise and judgment of its team” (FAI, About). Every position in this file is therefore a named scholar’s, quoted from FAI’s own publications, and is labelled as such.
• Overlaps with the evaluator map, tie by tie. Coefficient Giving is on FAI’s list and on AVERI’s funders list (AVERI, About); the Survival and Flourishing Fund is on FAI’s list and is credited at several AI Evaluator Forum members; the John Templeton Foundation is on FAI’s list and is a listed funder of Gillian Hadfield’s research on her CV; and four of the six companies from which AVERI says it has received API credits, Amazon, Anthropic, Microsoft and OpenAI, appear on FAI’s list of corporate supporters and event sponsors (AVERI, About; FAI, About).

Institutional Voice & Intent

FAI speaks as a builder-and-policymaker bridge: “We are builders, hackers, and scholars advancing an optimistic vision of the future—aligning technology with human ends to forge a more perfect union between innovation and the American republic” (FAI, About). Its scholars write in the register of drafters. One announced his draft as “my proposal for those incremental steps,” a proposal that “does not regulate the development of the technology itself and instead requires developers to tell the public how they are building the technology” (Dean Ball, Hyperdimensional). Another, on a House bill: “Essentially, the FRONTIER Act creates a free market of verifiers and auditors” (Daniel King, FAI Policy Gradients, 1 September 2026).

Stated intent: Tighten the coupling between innovators and policymakers: “Realizing this future requires far tighter coupling between innovators and policymakers than exists today” (FAI, About). Observed intent: The same, on the record: testimony and comments, a fellowship for a stated constituency, a litigation project, a 501(c)(4) and a stream of drafts. FAI’s Q3 2025 update: the AI Action Plan “reflects several recommendations made by members of the FAI team, including bolstering export controls on advanced AI chips and semiconductor manufacturing, expediting environmental permitting for data center and energy projects, unlocking federal lands in regions that can support large AI clusters, and strengthening oversight of frontier AI companies to avoid worst-case scenarios” (FAI, Q3 2025 update). Gap: None on direction. What the pages cannot show is how much of a revenue base that is 99.1 percent contributions comes from the labs and other companies on its sponsor list, since no amounts are published, or how the sponsor relationship bears on its scholars’ proposals for the labs. The record raises that and cannot settle it; for the structural reading it does not need to.

Interest: FAI states that it takes no institutional positions, so these are positions argued by its scholars, in their own words. For each: what adoption would pay FAI, as documented; the scholars’ best case; and who else gains from the same rule.

1. Position: federal preemption of state AI law, with developer transparency. Ball: “I have long been a supporter of a federal AI law with broad (though not universal) preemption of state regulations” (Hyperdimensional, The AI Patchwork Emerges). FAI’s Q2 2026 update: “Representatives Trahan and Obernolte released the Great American AI Act, a discussion draft structured like Dean’s “Be It Enacted” proposal” (FAI, Q2 2026 update). King, of the successor bill: it “preempts new state rules that cover auditing, verification, transparency, and incident reporting around catastrophic risks, while generally leaving states free to regulate how AI systems are deployed and used” (FAI, FRONTIER Act).
   • What adoption would pay it. A drafting credit and a place in the federal rule, which FAI reports itself; a 501(c)(4) affiliate “advancing energy and AI policy” with “government affairs efforts”; and, for the companies on its sponsor list, one national rule in place of state rules (FAI, Q1 2026 update; FAI, About). No payment to FAI for any rule is documented.
   • Its scholars’ best case. “A national framework is optimal on issues where a heap of state laws would create duplicative reporting burdens or—even worse—contradictory requirements” (King); Ball: “The rapaciousness with which states have sought to regulate AI is unprecedented for an emerging macroinvention” (FAI, FRONTIER Act; Hyperdimensional, “Be It Enacted”).
   • Who else gains from the same rule. Every developer now facing differing state rules; and the House authors of the draft. The opposition is documented in King’s own account: “More than 200 state legislators from 42 states signed a bipartisan letter opposing the preemption provision,” and that criticism produced a narrower preemption in the later bill.
2. Position: government-licensed independent verification organizations for the largest developers. King: “Instead of baking detailed technical standards into law, the bill licenses independent verification organizations to assess whether each very large frontier developer has a framework in place that, when followed properly, is adequate for reducing catastrophic risk.” Dillingham, on the House discussion draft: Ball “has advocated a marketplace of Independent Verification Organizations” (FAI, FRONTIER Act; FAI, Control Failure).
   • What adoption would pay it. A statutory buyer for the field in which FAI’s officers hold seats: each “very large frontier developer” must “retain a licensed IVO of its choice,” by King’s description of the bill; FAI’s Vice President of Growth co-founded and sits on the board of AVERI, an auditing institute; its President and CEO is a Fathom affiliate member and FAI is a Fathom partner (FAI, FRONTIER Act; AVERI, Team; Fathom, About; Fathom, Partnerships). No page read states that FAI applies or intends to apply to be an auditor, and none documents a fee to FAI.
   • Its scholars’ best case. King: “The bill’s concept for third-party technical experts to take ownership of technical decision-making makes it especially resilient to future changes in AI capabilities and company practices.” Dillingham: “we need regulatory institutions with real technical expertise that evaluate frontier labs” (FAI, FRONTIER Act; FAI, Control Failure).
   • Who else gains from the same rule. The evaluators and insurers that would be licensed; Fathom, whose sponsored California statute uses the same IVO term; and, by the scholars’ own account, the labs, which pick their verifier. Documented critique in the same literature: a coalition on the 2025 California bill wrote that “It is not clear whether there are any entities currently in the United States other than perhaps the sponsor of this legislation who might be able to comply with the current MRO qualifications in the measure” (Senate Judiciary analysis, 29 April 2025).
3. Position: voluntary pre-deployment testing by the government, not a mandatory “FDA for AI.” FAI’s Q2 2026 update, of an executive order: it “reflects the case Dean has made that the federal government should evaluate frontier models for catastrophic cyber and biological capabilities through voluntary pre-deployment agreements, rather than a mandatory “FDA for AI” licensing regime,” and records that the Commerce Department’s Center for AI Standards and Innovation signed agreements with Google DeepMind, Microsoft and xAI, “extending its voluntary pre-deployment reach to all five leading U.S. frontier labs” (FAI, Q2 2026 update).
   • What adoption would pay it. A voluntary regime leaves release timing with the developer, which is the position of the labs on its sponsor list; FAI’s fellow is now employed by one of them. No payment to FAI is documented.
   • Its scholars’ best case. Ball, of a “voluntary” order he says operated as a licensing gate: “nobody knows what the requirements are to get licensed” (Hyperdimensional, What Should Be Done). Dillingham, for the other side of the question: “any regime that clears each model before public release (such as the “FDA for AI” idea that some have floated) will not suffice” (FAI, Control Failure).
   • Who else gains from the same rule. Every lab that would otherwise face pre-clearance; the national-security agencies that keep testing for threats.
4. Position: a mandate where the scholars see consensus: nucleic-acid synthesis screening. The letter FAI’s President and CEO, Ball and a senior fellow led: “This is a rare moment of agreement across stakeholders that are often at odds,” signed by the chief executives of OpenAI, Anthropic, Google DeepMind and Microsoft AI, and, from FAI, Graves, Ball and others (screendna.org; FAI, Q2 2026 update). What adoption would pay it: nothing documented; the mandate falls on “providers of synthesized DNA and manufacturers of synthesis machines.” Who else gains: the signatories’ stated aim is a “consistent national standard rather than a patchwork of conflicting laws.”

Interest is a documented fact about each position, not a claim about motive, and not a verdict that the argument is wrong. Each best case is stated at the strength its author states it.

Position in the Apparatus

FAI is a policy organization beside the evaluators, with three documented seats in the coalition. Fathom sponsored California’s statute designating independent verification organizations and lists FAI among its partners; AVERI argues for mandatory audits and has FAI’s Vice President of Growth on its board; and Ball, FAI’s fellow, writes the proposals for both and now works at OpenAI. Its nearest peers on the map are GovAI, which supplies research and people to government rather than text, and the Center for AI Safety, whose Action Fund sponsored a state bill. FAI neither sponsors a statute nor tests a model.

Actions & Leadership Choices

CONDUCT: WATCH — AN OPEN-VOICE THINK TANK WHOSE FELLOWS DRAFT THE RULES, WITH LAB SPONSORS, A BOARD SEAT AT AN AUDIT INSTITUTE AND PARTNER TIES TO THE DESIGNATION SPONSOR, ALL DISCLOSED. No wrongdoing is alleged. FAI publishes its supporters, its gift policy, its filing and its “open voice” policy, and its fellow publishes his own affiliations with each move. The structural point is the one the evaluator map makes about every actor in it: disclosure of ties, not absence of them. What is particular to FAI is that its scholars write the preemption and verifier texts, its sponsor list includes the labs those texts bind, and its officers sit in the audit institute and designation-sponsor networks, with no filing yet showing how much of its money comes from whom.

• 2014 to 2023. Founded as Lincoln Labs / Lincoln Network; refounded as FAI in 2023 (FAI, About).
• Summer 2025. Launches the Conservative AI Policy Fellowship; says “two fellows” took “AI roles at the White House” (FAI, Q3 2025 update).
• 11 August 2025. Ball, “returning to the private sector” from OSTP, rejoins FAI, as MeriTalk reported (MeriTalk).
• 2 October 2025. Ball publishes the preemption proposal “Be It Enacted” (Hyperdimensional).
• 12 November 2025. FY2024 Form 990 filed (ProPublica Nonprofit Explorer).
• Q1 2026. Werner joins as Vice President of Growth; FAI Action launches in March; FAI files an amicus brief in Anthropic v. U.S. Department of War (FAI, Q1 2026 update).
• June 2026. Graves, Ball and a senior fellow lead the nucleic-acid screening letter; FAI announces Ball’s move to OpenAI on 18 June; launches Frontier Legal Defense on 29 June, a “rapid-response legal team that combats the concentration of power, incumbent rent-seeking, and government overreach in AI” (screendna.org; FAI, 18 June 2026; FAI, 29 June 2026).
• 23 July and 1 September 2026. FAI scholars publish “Control Failure” and “The FRONTIER Act Is Congress’s Best AI Bill Yet” (FAI, Control Failure; FAI, FRONTIER Act).

Alliance Map

Where ties on different axes run at the same time, the overlap is the finding. In the same quarter, FAI lists the labs as sponsors, a fellow joins one of them, another fellow’s bill text is followed by a House draft, its Vice President of Growth sits on the board of the auditing institute whose paper that fellow co-authored, its President and CEO is on the designation sponsor’s affiliate list, its General Counsel files amicus briefs for one lab, and its former board member directs the White House office in which the fellow served. The documents show disclosed seats, listings and sponsorships, not coordination, and no designation has been made.

Reach Assessment

Institutional: Moderate. FAI cannot require anything of anyone. Its reach is drafts that members of Congress release in its fellow’s structure, testimony and comments, a fellowship that placed graduates in the White House, a former board member who directs the White House science office, and a seat at the evaluator coalition’s boards and partner lists; all of it influence by invitation.

Memetic: Moderate. “Be It Enacted,” the preemption proposal, and the “marketplace of Independent Verification Organizations” travel in policy discussion under its scholars’ names; FAI’s quarterly updates list its scholars’ bylines and quotations in national outlets.

Civilizational: Low to moderate, structural. FAI tests no model and holds no license. If the House bills it commends were enacted, its scholars’ texts would sit upstream of both the preemption of state AI law and the licensing of verifiers of the labs that sponsor it; that is the seam, and the sponsor list, the board listings and the fellowship ties are on pages FAI itself publishes. The reach is the finding; the malice is not asserted.

Score basis. Atk 6 is reach: FAI’s own updates say “the administration’s AI Action Plan reflects several recommendations made by members of the FAI team,” that a House discussion draft is “structured like Dean’s “Be It Enacted” proposal,” and that “two fellows” of its fellowship took “AI roles at the White House” (FAI, Q2 2026 update; FAI, Q3 2025 update). It is not higher because no statute or procurement rule names it, it holds no gate or designation power, and these accounts are its own. It sits one above AVERI (5/6/4, hazard 50), whose reach rests on a coauthored standard and an endorsed bill that became law, and one below Fathom (7/7/6, hazard 66.7), whose sponsored statute passed and was chaptered; two below GovAI (7/8/7, hazard 76), which supplies the government’s institutes with research and people; and three below the Center for AI Safety (9/7/6, hazard 83), whose output labs sign. Def 5 is funding insulation: FAI names about two dozen foundations and about forty companies, among them OpenAI, Anthropic, Meta, Alphabet, Microsoft and Amazon, as supporters, and its FY2024 contributions were $5,106,107, or 99.1 percent of revenue (FAI, About; ProPublica Nonprofit Explorer). The points above 3 are the breadth of the base, the published gift policy and a return on file; the points off are that no amounts or donor shares are published and that it takes sponsorship from the labs whose rules its fellows draft. AVERI holds 6 because it says it accepts no cash from frontier AI companies; Fathom holds 7 because it takes no corporate donations. Hp 5 is entrenchment: an IRS exemption from March 2015, an eleven-year record, a 2023 refounding, a headcount FAI says nearly tripled, a new Capitol Hill headquarters and a 501(c)(4) affiliate (ProPublica Nonprofit Explorer; FAI, Q2 2026 update). The points off are that no statute, contract or program of record installs it, net assets were $3,427,251 against expenses of $3,719,664, and research and commentary can be ignored. Peers on hp: AVERI (4) has a 15-month record; Fathom (6) has two chaptered statutes and an order with dated deadlines. The values hold at 6/5/5, hazard 53.3.

Sources: FAI — About; FAI — Conservative AI Policy Fellowship; FAI — Zach Graves; FAI — Grace Werner; FAI — Dean Ball; FAI — Dean Ball Joins OpenAI, 18 June 2026; FAI — Frontier Legal Defense, 29 June 2026; FAI — Q3 2025 update; FAI — Q4 2025 update; FAI — Q1 2026 update; FAI — Q2 2026 update; FAI — The FRONTIER Act Is Congress’s Best AI Bill Yet; FAI — Control Failure; ProPublica — The Foundation For American Innovation; AVERI — Team; AVERI — About; AVERI — Frontier AI Auditing; Fathom — About; Fathom — Partnerships; screendna.org — open letter; Hyperdimensional — “Be It Enacted”; Hyperdimensional — The AI Patchwork Emerges; Hyperdimensional — Leviathan Waking; Hyperdimensional — What Should Be Done; Hyperdimensional — On Private Governance; MeriTalk — Dean Ball Departs OSTP; Senate Judiciary Committee analysis of SB 813, 29 April 2025.