AI Policy Institute (AIPI)

Theory of Change

AIPI's theory of change operates through three steps: (1) conduct public opinion polls showing Americans want AI regulation, (2) disseminate those polls through media and direct policymaker engagement to create a "permission structure" for action, (3) build political will for legislation.

Founder Daniel Colson articulates the core argument in democratic terms: "if the tech companies that are creating and deploying and managing those technologies aren't democratic institutions, then to what extent is America democratic?" He reads Bostrom's Superintelligence in 2014 as his origin story and frames AI risk through a civilizational collapse lens -- comparing it to European destruction of Amazonian civilizations.

The c3 (AIPI) produces the polling. The c4 lobbying arm (AI Policy Network / AIPN) takes that data to Congress. AIPN's issue page is more explicitly x-risk focused, quoting lab CEO timelines on AGI (2-5 years) and discussing loss-of-control scenarios. The two entities together form a polling-to-lobbying pipeline.

Notably, AIPI does not produce policy research, technical analysis, or model legislation. Its entire intellectual output is polling data and press releases. Other organizations (IAPS, CAIP, CAIS) provide the policy substance; AIPI provides the political mandate.

What They Do

Polling. AIPI has conducted 20+ polls with YouGov since August 2023. Key findings are remarkably consistent: 72% want to slow AI development, 82% don't trust tech executives to self-regulate, 83% believe AI could cause catastrophic events, 80% prefer mandatory safety measures over self-regulation. These numbers hold across party, geography, and time.

The methodologically strongest output was the September 2024 collaborative poll with Dean Ball (Mercatus Center, AI regulation skeptic). Despite jointly designed questions presenting both pro and con arguments, 62% of California voters supported SB 1047. This preempted push-poll criticism.

SB 1047 engagement. AIPI ran three sequential polls showing 59-70% CA voter support for SB 1047, the California AI safety bill. These polls became part of the legislative record and were cited by Wikipedia alongside the California Chamber of Commerce's push poll (described as "badly biased" by the American Prospect). Newsom vetoed SB 1047 regardless.

Federal lobbying. AIPN spent $349K on federal lobbying in the first three quarters of 2025 and held 24+ private events with members of Congress. Their outside lobbying firm, Skyline Capitol, targets intelligence and surveillance committees specifically. Focus areas: AGI preparedness, nonproliferation, testing protocols, national security readiness.

Anti-preemption fight. AIPI/AIPN positioned themselves in the bipartisan coalition opposing federal preemption of state AI regulation. AIPN's Mark Beall was quoted in NBC alongside Steve Bannon and Elizabeth Warren. IFS/YouGov polling (with which AIPI collaborated) showed Americans reject preemption 3-to-1. The NDAA preemption attempt was defeated 99-1 in the Senate.

Continued 2026 activity. A February 2026 poll of 3,969 likely voters in battleground districts found 81% prefer regulatory guardrails vs. 6-8% favoring no regulation.

Key People

Daniel Colson -- Co-Founder and Executive Director of both AIPI and AIPN. The org is essentially his vehicle. Background: sociology researcher at Leverage Research (a Bay Area organization with documented cult-like characteristics), co-founder of Reserve (crypto fintech with Sam Altman and Peter Thiel as investors), co-directed EA Build (CEA project, 2016). He distanced from EA around 2020, believing the movement was "co-opted by technologists" ignoring government regulation. Funds Amazon lidar archaeology from personal wealth. His worldview combines technological determinism with civilizational collapse thinking.

Mark Beall, Jr. -- President of Government Affairs at AIPN. Former inaugural Pentagon AI Policy Director at DoD Joint AI Center, co-authored frontier AI risks report for State Department. The most credentialed DC person in the operation.

Peter Wildeford -- Head of Policy at AIPN. Co-founder and former CSO of IAPS, co-founded Rethink Priorities. His move from IAPS to AIPN is a notable personnel cross-pollination, bringing AI governance expertise and forecasting methodology into the lobbying operation.

Combined team across both entities: ~10 named staff. AIPN has 8 named staff with serious DC experience (former Hill staffers from Pelosi, Schumer, and Blumenthal offices; Pentagon alumni; former Meta policy team). AIPI's about page lists only Colson.

Money and Incentives

Funding opacity is AIPI's defining financial characteristic. Every comparable AI policy org has identifiable major funders (IAPS: $15.3M from OP; CAIS: $12M+ from Moskovitz; CAIP: SFF). AIPI has zero Coefficient Giving grants, zero identified institutional funders, and zero 990 filings available. Its only public funding description: "initially funded by anonymous donors from tech and finance industry" (Semafor, Oct 2023).

AIPN budget (from Manifund page, 2025): $1.17M total need. $629K received + $125K pledged from individual donors. $412K gap. Allocation: ~65% staff, ~20% outside lobbying firms, ~15% ops/compliance. AIPN also sought $100K on Manifund (an EA-adjacent platform).

Estimated AIPI costs: YouGov polls typically cost $5K-$15K each for ~1000 respondents. At 20+ polls, polling costs alone are $100K-$300K. Add PR firm (Slingshot Strategies) and overhead, and AIPI's annual budget is likely $300K-$500K minimum.

No AI company money identified. No lab affiliations, no compute dependencies, no industry board seats. AIPI appears genuinely free of the industry funding conflicts that affect many governance orgs.

Self-funding possibility. Colson has personal wealth from Reserve (investors included Altman and Thiel) and funds archaeology expeditions personally. He may partially self-fund AIPI operations.

Key incentive question: The anonymous donors from "tech and finance" are the unknown. Without knowing who funds AIPI, it is impossible to fully evaluate the org's incentive structure.

What Others Say

American Prospect validated AIPI's polling methodology relative to the California Chamber of Commerce's "badly biased" push poll on SB 1047, but also characterized AIPI as "a pro-regulation polling shop" -- correctly identifying that AIPI has a regulatory agenda even if its polls are methodologically sound.

The Dossier describes the broader ecosystem AIPI operates in as "The AI Safety Racket" -- a "dark money" network of EA-funded groups using child safety framing and bipartisan cover to push California-style regulation into red states. AIPI is not directly named, but the ecosystem is described as deceptive astroturfing backed by Moskovitz.

Reason Magazine calls EA-backed AI regulation "the authoritarian side of effective altruism" -- framing AIPI's allies (CAIP, CAIS) as seeking dictatorial agency powers. AIPI's polling supported the bills being criticized.

American Affairs Journal offers the most intellectually serious counterargument: "Beyond Safetyism" argues AI regulation should focus on demonstrable present harms (education, children, emotional manipulation) rather than speculative x-risk. The author argues the "boy who cried wolf" dynamic makes catastrophic risk framing politically counterproductive.

IFS polling (independent) corroborates AIPI's core findings with a 6,200-respondent survey: 71% hold negative views of AI, 80% want companies held liable for child harms, and pro-regulation candidates have electoral advantages across party lines. Some respondents accused IFS's survey of being "pro-AI" -- suggesting public AI hostility is even deeper than polls capture.

AI safety community: no engagement. Zero EA Forum/LessWrong discussion of AIPI. The org operates entirely outside the rationalist discourse.

What's Absent

No identifiable institutional funders despite 2.5 years of operation. No board of directors disclosed. No 990 filings available. No policy papers or research reports (only polls). No technical AI safety engagement. No annual report or impact assessment. No substantive policy proposals (unlike CAIP's model legislation or IAPS's compute verification). No engagement from AI safety researchers. AIPI's "co-founder" (besides Colson) is unnamed.

The deepest absence: no evidence of a causal chain from AIPI's polling to actual policy outcomes. SB 1047, AIPI's most prominent engagement, was vetoed. Whether AIPI's polling has influenced any legislator's vote or any bill's passage is undocumented.