Center for AI and Digital Policy (CAIDP)

Theory of Change

CAIDP aims to establish democratic governance of AI through four mutually reinforcing mechanisms:

Scoring countries on AI governance using the annual AI and Democratic Values Index (AIDV), benchmarking 80 countries against 12 metrics derived from the OECD AI Principles, UNESCO Recommendation on AI Ethics, and the Universal Declaration of Human Rights.
Filing regulatory complaints that force enforcement of existing consumer protection and privacy law against AI companies. The March 2023 FTC complaint against OpenAI (46 pages, requesting a halt to GPT-4 deployment) is the primary example.
Training future AI policy practitioners through semester-long policy clinics, now with 1,500+ graduates from 121 countries. Clinic participants produce country reports for the AIDV Index, creating a virtuous cycle.
Advising international bodies on AI governance frameworks -- OECD, Council of Europe, UNESCO, G20. CAIDP was an official observer in CoE AI Treaty negotiations (signed by 41 countries, September 2024).

Founder Marc Rotenberg, in the Carnegie Council podcast: "What seemed most important to me was to establish a system of democratic governance for this new technology... we are trying to promote democratic governance, we are trying to promote the rule of law, and ensure the protection of fundamental rights."

The intellectual foundation is the 2018 Universal Guidelines for AI, which Rotenberg helped draft. These include a "termination obligation" -- if an AI system loses human control, whoever deployed it must take it down. Rotenberg explicitly connects this to existential risk concerns.

What They Do

AIDV Index (flagship): Published annually since 2020. Grown from 30 countries to 80. The 2025 edition runs 1,500 pages with 7,000+ footnotes and 1,000+ contributors from 120 countries. Top-ranked: Canada, Japan, Korea, Netherlands, UK. The US sits in the middle tier. Uses 5 tiers across 12 metrics. AI World EU called it "the bible of AI and democracy."

FTC complaint against OpenAI (strongest demonstrated impact): Filed March 30, 2023. Argued GPT-4 is "biased, deceptive, and a risk to privacy and public safety" under Section 5 of the FTC Act. Requested: halt deployment, independent assessment, incident reporting, rulemaking for generative AI. The FTC opened an investigation of OpenAI in July 2023. Carnegie Council host Wendell Wallach: CAIDP was "the most visible" force prompting FTC action. (An O'Melveny analysis notes the FTC likely lacked authority to halt deployment, but could investigate.)

Policy clinics: 1,500+ applicants for Fall 2025 alone. Participants from 120+ countries study OECD, UNESCO, and other frameworks, then produce country reports feeding the AIDV Index. A first-person review described a "warm learning environment" with professionals from 17+ nations.

Council of Europe AI Treaty advocacy: CAIDP was an official observer in treaty negotiations. The treaty was adopted May 2024 and signed by 41 countries in September 2024. Both Rotenberg and Hickok are listed as endorsers. CAIDP is now running a ratification campaign. (Civil society groups, including ECNL, warned that states excluded civil society observers from the actual drafting group -- limiting CAIDP's direct influence on the text.)

Other: Congressional testimony (Hickok before House Oversight, March 2023). Ban Facial Surveillance campaign. FOIA Transparency Project. Amicus briefs (Gonzales v. Google). Comments to OSTP under both Biden and Trump administrations. Newsletter with 90,000+ subscribers; LinkedIn following of 85,000+ (#2 behind Stanford HAI among AI policy organizations).

Key People

Marc Rotenberg -- Founder and Executive Director (age 65). Harvard AB, Stanford JD, Georgetown LLM. Co-founded EPIC in 1994, led it until forced departure in April 2020 (COVID workplace incident -- he continued going to work after learning he might have COVID-19). Subsequently filed lawsuits against EPIC (dismissed by stipulation) and Politico (dismissed for lack of jurisdiction, with a Volokh Conspiracy analysis finding most libel claims "pretty weak"). Georgetown Law adjunct since 1990. 60+ Congressional testimonies. Member of OECD Expert Group on AI. Three-time DC chess champion. Rotenberg IS the organization -- his personal networks, four decades of policy work, and institutional relationships are CAIDP's primary asset. Compensation: $76,500 (2024).

Merve Hickok -- President and Policy Director. Founder of AIethicist.org. University of Michigan lecturer. CFR-Hitachi Fellow. UNESCO AI Ethics Expert. Council of Europe AI treaty observer. Testified before US Congress and Turkish National Assembly. "100 Brilliant Women in AI Ethics." Previously 15+ years in Fortune 100 companies. Brings international AI ethics credentials and policy expertise. Compensation: $79,000 (2024).

Board appears to have only one independent member: Pablo Molina (Treasurer, unpaid) -- CISO at Drexel University, Georgetown Law lecturer.

Money and Incentives

Revenue: $115K (2021) to $258K (2022) to $714K (2023) to ~$924K (2024). 8x growth in 4 years. Total assets reached $849K in 2024.

Funding sources: Craig Newmark Foundation, Patrick J. McGovern Foundation ($500K confirmed in 2024), Fund for Constitutional Government, Planet Heritage Foundation. Zero funding from Coefficient Giving/Open Philanthropy. Zero funding from any AI lab or tech company.

Business model: 100% philanthropic grants from digital rights foundations. Zero program revenue, zero investment income until 2024 ($24K). Zero fundraising expenses -- relies entirely on Rotenberg's personal relationships.

Compensation: Rotenberg $76.5K, Hickok $79K, plus $81K in other wages (2024). Total compensation of ~$237K on ~$924K revenue is remarkably lean. These salaries are far below market for the expertise deployed.

Independence: CAIDP exists in the "digital rights philanthropy" ecosystem, completely separate from the "x-risk philanthropy" ecosystem dominated by Coefficient Giving, Jaan Tallinn, and tech billionaires. This means zero financial ties to AI labs, zero compute dependencies, and genuine independence from the entities CAIDP criticizes -- but also exclusion from the largest pool of AI safety funding.

Concentration risk: The McGovern Foundation's $500K alone is more than half of likely 2024 revenue, creating significant funder concentration.

Incentive alignment: Low salaries, no commercial revenue, no lab ties, and no equity stakes suggest strong mission alignment. The primary incentive concern is not financial corruption but institutional conservatism -- CAIDP's funding comes from traditional foundations that may prefer incremental governance over the kind of emergency response that existential risk might warrant.

What Others Say

InfluenceWatch describes CAIDP as "supporting slowing AI development with an emphasis on legal guidelines." Factually accurate but framings differ -- CAIDP would say "governance," not "slowing."

DAIR Institute (Timnit Gebru et al.) criticized the Pause Giant AI Experiments letter that Rotenberg signed: "Those hypothetical risks are the focus of a dangerous ideology called longtermism that ignores the actual harms resulting from the deployment of AI systems today." CAIDP sits at an unusual intersection -- rooted in the rights/ethics tradition but willing to engage with slowdown/pause arguments that most ethicists reject.

Civil society concern: ECNL warned that Council of Europe states excluded civil society observers from the AI convention's drafting group, despite CAIDP's official observer status. This suggests limits to CAIDP's "inside" advocacy approach.

Academic critique of AIDV methodology (paywalled): concerns about "lack of granularity in index scoring" and that the index "measures policy commitments more than implementation." This is a substantive issue -- scoring whether a country adopted an AI strategy is different from scoring whether that strategy reduces actual AI risk.

No engagement from x-risk community: Zero mentions on LessWrong, EA Forum, or Alignment Forum. No 80,000 Hours interview. No Coefficient Giving funding. CAIDP and the x-risk governance ecosystem (GovAI, CLTR, etc.) operate in parallel without interaction. Whether this represents principled independence or a missed collaboration opportunity is debatable.

What's Absent

No independent effectiveness assessment. Despite CAIDP measuring other countries' AI governance, no one has measured CAIDP's own impact beyond the FTC complaint.
Minimal board oversight. Three known officers/board members, two of whom are the primary staff. One independent board member (unpaid).
No technical AI safety capacity. Zero engagement with alignment research, model evaluations, dangerous capability assessments, or frontier safety frameworks (RSPs, etc.).
No succession plan. CAIDP depends almost entirely on Rotenberg (age 65) and his personal networks. Key person risk is extreme.
No engagement with AI labs beyond adversarial actions. No evidence of dialogue with Anthropic, DeepMind, Meta AI, or the AI Safety Institute.
CAIDP.org inaccessible -- returns 403 errors, preventing verification of claims. Ironic for a transparency advocacy organization.