Skip to the page content
Passes our checks

Center for Artificial Intelligence Safety Inc

EIN 88-1751310 · San Francisco, CA

Public records show a registered, tax-exempt organization in good standing.

LegitimacyPassesRegistration and standingFinancial healthGood85 out of 100GovernanceStrong93 out of 100Impact evidenceDescribes programsHow well its results are evidenced

What it says it does

to reduce societal-scale risks associated with artificial intelligence AI by advancing safety research, building the field of AI safety researchers, and promoting safety standards.

Income
$10.2 million
Spending
$7.2 million
Spent on programs
88%
Reserves
18.3 months of spending
Largest source of income
donations and private grants, 94%
Employees
25
Volunteers
2
Highest pay
$321,283, executive dir.

Source: Form 990 for the year ending December 31, 2024, filed with the Internal Revenue Service. Rating method 2026.2. How we rate

Financial health

Good: 85 out of 100. Higher than 68% of charities in its peer group (science, $10m to $100m).

  • Reserves

    35 of 35

    Spendable reserves would cover 18.3 months of spending.

  • Debt

    20 of 20

    It owes 8% of what it owns.

  • Living within its means

    15 of 15

    Income exceeded spending by 30% of income.

  • Income trend

    0 of 15

    Income fell 36% on the year before.

  • Spending floor

    15 of 15

    88% of spending went to programs and raising each dollar of donations cost 2 cents.

Source: Form 990 for the year ending December 31, 2024, filed with the Internal Revenue Service. Rating method 2026.2. How we rate

Governance

Strong: 93 out of 100. Higher than 35% of charities in its peer group (science, $10m to $100m).

  • Independent board

    18 of 25

    2 of 3 voting board members are independent.

  • Conflict of interest policy

    15 of 15

    Has a written conflict of interest policy.

  • Whistleblower policy

    8 of 8

    Has a written whistleblower policy.

  • Record keeping policy

    7 of 7

    Has a written policy on keeping and destroying records.

  • Board minutes

    5 of 5

    Keeps minutes of board meetings.

  • Board sees the tax return

    10 of 10

    Gave the board a copy of this return before filing it.

  • Pay set independently

    5 of 5

    Sets its chief executive's pay through independent review and comparison.

  • Independent check of accounts

    15 of 15

    Its accounts were audited by an independent accountant.

  • No reported misuse

    10 of 10

    Reports no diversion of assets, excess benefits or loans to insiders.

Source: Form 990 for the year ending December 31, 2024, filed with the Internal Revenue Service. Rating method 2026.2. How we rate

Impact evidence

Describes programs. Its tax return describes what it does but we have found no measurable results in it.

Programs described in its return (4)
  1. $2.9 million spent. Compute Cluster: CAIS runs an initiative to provide free compute for research projects in ML safety, based on a cluster of 256 A100 GPUs with dedicated support staff to enable non-industry researchers to conduct advanced AI safety research requiring cutting-edge models. Since launching in 2022, the CAIS Compute Cluster has become a cornerstone of global AI safety research. By 2024, more than 400 researchers had used the cluster to conduct safety-focused ML research, generating 121 research papers and accumulating over 5,000 citations. The cluster continues to remove financial and technical barriers for non-industry researchers, enabling cutting-edge experimentation through free access to compute.
  2. $1.2 million spent. Research: CAIS conducts foundational research aimed at mitigating societal-scale risks posed by artificial intelligence. We develop technical benchmarks and methods which improve AI safety. We also examine AI safety from a multidisciplinary perspective that incorporates safety engineering, complex systems, international relations, and other disciplines. In 2024, CAIS made major strides in foundational AI safety by developing new tools to measure and mitigate national-security risks posed by advanced models. This work included releasing the WMDP Benchmark, the first public dataset for evaluating hazardous knowledge in AI systems; creating circuit breakers, a defense mechanism that reliably prevents harmful outputs even under adversarial attacks; and coordinating Humanity's Last Exam, a global collaboration of nearly 1,000 contributors from more than 500 institutions across 50 countries to assess frontier-level reasoning across 3,000 expert-level questions. CAIS researchers also published 8 papers over the year.
  3. $777,874 spent. Advocacy: CAIS works to bring AI safety research into practice through engagement with industry leaders, policymakers, and research labs, aiming to build awareness and establish standards for safe AI development and deployment. CAIS saw significant growth in public engagement, expanding its AI Safety Newsletter to 24,000 subscribers, a nearly threefold increase from the prior year, and contributing to over 100 media stories in outlets such as The New York Times, The Wall Street Journal, and The Washington Post. Through research translation, public communication, and collaboration with academic and industry partners, CAIS helped elevate AI safety as a mainstream concern and provided technical grounding for ongoing policy discussions at both state and federal levels.
  4. $520,385 spent. Field Building: CAIS builds and strengthens the AI safety research community through various initiatives designed to provide support and lower barriers to entry. In 2024, CAIS expanded its field-building initiatives by making AI safety education more accessible and interdisciplinary. We published the textbook Introduction to AI Safety, Ethics & Society, and launched an accompanying online course that enrolled 240 students from varied academic backgrounds. CAIS also strengthened community engagement by hosting workshops at major conferences including NeurIPS, where our workshop on agentic AI systems accepted 34 papers, and organizing AI safety socials at ICML and ICLR with 200+ researchers participating. The SafeBench competition, supported by Schmidt Sciences, attracted nearly 90 benchmark submissions to advance empirical AI safety evaluations.

Source: Form 990 for the year ending December 31, 2024, filed with the Internal Revenue Service. Rating method 2026.2. How we rate

Legitimacy checks

  1. Is it registered with the IRS? Passed.

    Listed in the IRS master file of tax-exempt organizations as a 501(c)(3) charitable organization.

    Source: Exempt Organizations Business Master File, Internal Revenue Service, file dated October 13, 2026.

  2. Has the IRS revoked its tax-exempt status? Passed.

    Not on the IRS automatic revocation list.

    Source: Automatic revocation of exemption list, Internal Revenue Service, file dated November 6, 2026.

  3. Are donations tax-deductible? Passed.

    On the IRS list of organizations eligible to receive tax-deductible contributions.

    Source: Publication 78 data: organizations eligible for tax-deductible contributions, Internal Revenue Service, file dated October 13, 2026.

  4. Is it keeping up with its IRS filings? Passed.

    Most recent return on file covers the year ending December 2024.

    Source: Exempt Organizations Business Master File, Internal Revenue Service, file dated October 13, 2026.

  5. How long has it been tax-exempt? Passed.

    Recognized by the IRS since January 2023.

    Source: Exempt Organizations Business Master File, Internal Revenue Service, file dated October 13, 2026.

  6. Is it in good standing with state regulators? Passed.

    Not on California's list of charities barred from soliciting. Other states are not yet checked.

    Source: May Not Operate or Solicit for Charitable Purposes list, California Attorney General, Registry of Charities and Fundraisers, file dated November 4, 2026.

Ratings and checks are our reading of public records, using the published method. They are not a statement about anyone's honesty. If you work for this organization, you can send corrections or results, confirm you want gifts through Veridonate, or ask to be removed from giving.