Americans for Responsible Innovation (ARI)

Theory of Change

ARI's theory of change rests on two core claims:

First, near-term AI harms and long-term existential risk are causally connected. In Carson's words: "These near-term harms, whether it's algorithmic bias and discrimination, dangers to children, AI psychosis... those are very related to concerns about the degradation of democracy and our ability to think clearly about the future. As the institutions become degraded, they'll be harder to use... to grapple with misaligned AI" (Roll Call, Jan 2026).

Second, policymakers should build "regulatory muscle" -- people, institutions, processes, and laws -- while deliberately suspending judgment on which AI risk scenario is correct. Their first principle: "The only thing we can be certain of is that we cannot say with high confidence what trajectory the technology will take... A key challenge for policymakers will be to suspend judgment of which prediction is 'right' and instead craft policies that can work in a variety of scenarios" (ARI Values page).

The mechanism: ARI lobbies Congress for specific legislation (AISI authorization, whistleblower protections, export controls, incident reporting, AI liability frameworks), builds coalitions with industry and civil society, runs grassroots campaigns, and supports pro-regulation candidates through PAC activity. They explicitly aim to counter Big Tech's lobbying advantage by providing independent, non-corporate-funded expertise to policymakers.

What They Do

ARI is a full-spectrum Washington advocacy operation: lobbying, coalition building, grassroots mobilization, policy research, and electoral engagement. Three activity types dominate:

Legislative wins. ARI helped pass the TAKE IT DOWN Act (landmark federal AI safety law on non-consensual deepfakes, signed by Trump, 2025) through coalition building, legal memos, and grassroots activation. They led the campaign defeating a 10-year state AI law moratorium buried in a reconciliation bill -- 25,000+ petition signatures, letter from 260 bipartisan state lawmakers, original research on consequences -- resulting in a 99-1 Senate vote to strip the provision.

Coalition lobbying. ARI co-led (with ITI) a coalition of 60+ organizations calling for AISI legislation at NIST. They participated in FAS policy sprints, co-led NIST funding advocacy with FAS, and built coalitions for AI whistleblower protections.

Policy research. Produced reports on agentic AI liability (three policy approaches), AI incident reporting systems, AI safety research highlights, preemption analysis, and Morning Consult polling on voter attitudes toward AI. Their most technically sophisticated output was a 2025 survey of misalignment research, evaluation awareness, CoT monitoring, and interpretability. No peer-reviewed original research.

Congressional testimony. Carson testified before Senate Commerce Committee on Section 230 (March 2026), arguing generative AI outputs are not third-party content and AI preemption would repeat Section 230's structural mistakes. He introduced the "meta-law" concept -- a law that determines who governs rather than how, becoming nearly impossible to correct once the industry matures.

Export controls. Active on Nvidia chip diversion, H20 chip restrictions, GAIN AI Act, and broader AI export control frameworks.

Key People

Brad Carson, President. Former Congressman (D-OK, 2001-2005), Under Secretary of the Army, President of University of Tulsa (left May 2025 to lead ARI full-time). Rhodes Scholar, Cherokee Nation member, Bronze Star. An unusual profile for AI policy -- his background is politics and military, not tech or EA. Also co-founded Public First Action (bipartisan super PACs, $20M from Anthropic, targeting $50M+).

Eric Gastfriend, Executive Director. Former CEO of DynamiCare Health (addiction telehealth). Harvard MBA, co-founded Harvard EA student group. Was "entirely self-funding" ARI in its earliest phase. $253K compensation (2024). Public-facing presence is minimal; Carson is the external spokesperson.

Team pattern. ~27 staff, overwhelmingly drawn from Congressional offices and government agencies (Klobuchar, Manchin, BIS, DOD, USAID, Heritage Foundation). A few staff have AI safety research backgrounds (former MATS researchers, former CAIS researcher, former GCRI deputy director). The highest-paid employee is SVP of Government Affairs Doug Calidas ($438K), not the Executive Director -- signaling that Hill relationships are the most valued asset.

Money and Incentives

Revenue. 2023: $502K (founding year). 2024: $5.24M (10x growth). Entirely grants and donations -- $0 in program services or product revenue.

Funder concentration. Open Philanthropy/Coefficient Giving provided $3M over two years (Aug 2024), roughly 57% of 2024 revenue. Other organizational funders: Inclusive Abundance Action, Omidyar Network. CRI funders: Founders Pledge (recommends for giving), Laidir Foundation, Packard Foundation, Sentinel Bio. Individual donors include both co-founders and Steve Newman.

No corporate funding policy. ARI explicitly refuses corporate or foreign government money. This limits revenue but strengthens independence claims.

The Anthropic question. Carson separately co-founded Public First Action, which received $20M from Anthropic (Feb 2026). ARI subsequently defended Anthropic in the Pentagon supply-chain dispute, calling the designation "a dangerous precedent." Both actions may be principled -- the Pentagon's designation was widely criticized -- but Carson simultaneously leads ARI and Public First Action, and no disclosed recusal policy exists. If Anthropic's commercial interests ever diverge from the public interest in AI safety, the lack of formal separation becomes a problem.

PAC activity. ARI PAC: $205K raised (2026 cycle). ARI Carey Committee: $92K raised, $60K to candidates (58% R, 42% D) in 2023-2024.

Competitive context. Big Tech spends $36M per half-year on AI lobbying (8 companies). OpenAI-backed Leading the Future PAC raised $125M+. Public First Action (Carson's separate entity) targets $50-75M but currently has ~$20M from Anthropic. ARI itself, at $5M, is outspent 5:1 to 10:1 by industry opponents. A congressional staffer: "by far the best at it are the tech groups."

Business model. Pure philanthropy. No revenue diversification. ARI is entirely dependent on a small set of EA-adjacent and progressive funders.

What Others Say

Semafor (Oct 2023): Placed ARI in context of EA-linked AI policy groups flooding DC. Critical conclusion: "lawmakers will need to also consider viewpoints from outside this bubble if they want to be effective."

TechPolicy.Press (May 2024): Called ARI and IAPS "better positioned than CAIP for moderate bills" but questioned whether the AI safety community can move from vague concerns to mobilizable political movement.

Reason (libertarian, July 2024): Critiqued EA-backed AI regulation as authoritarian paternalism. Not ARI-specific, but directly challenges the premise that regulation serves innovation.

David Sacks (Nov 2025): Accused EA-funded orgs of manufacturing AI distrust through a "$1 billion Doomer Industrial Complex." Sacks is the White House AI and crypto czar, which makes this the most powerful institutional opposition ARI faces.

Founders Pledge: Recommends CRI for giving, citing its "big tent" bipartisan approach, strong leadership, and early demonstrated impact. No weaknesses identified.

No direct criticism of ARI exists. The org is too new and small to have attracted dedicated detractors. All criticism is at the ecosystem level (EA influence in DC, "doomer" narrative).

What's Absent

ARI does not publish its board of directors. No CRI (501c3) financial data is available. There is zero engagement with the EA Forum, LessWrong, or Alignment Forum intellectual community (0 posts). No original technical research exists. No disclosed conflict-of-interest or recusal policy addresses the Carson-Anthropic-Public First Action nexus. No position has been stated on AI moratoriums/pauses, open-source AI, or domestic compute governance. No annual report or systematic impact metrics are published. Eric Gastfriend's personal views on AI risk are not publicly documented despite being the EA-connected co-founder.