Resource guide

Center for AI Safety: Who Runs It and Who Funds It

The nonprofit behind the 2023 AI extinction-risk statement and California's SB 1047 bill, which would have regulated frontier AI models.

Last updated October 22, 2026 1834-word guide By Jonathan West

The Center for AI Safety (CAIS) is a San Francisco nonprofit that works to reduce societal-scale risks from artificial intelligence (AI). Dan Hendrycks, its Executive and Research Director, leads it. It is registered as a 501(c)(3) nonprofit, and it has a separate 501(c)(4) sister group, the CAIS Action Fund, which sponsored a California AI bill. CAIS doesn't publish a donor list.

What the Center for AI Safety Is and Who Pays for It

The Center for AI Safety is a 501(c)(3) nonprofit run by Dan Hendrycks. Its most recent filing shows $10,235,085 in total revenue for fiscal 2024. It doesn't publish a donor list, and its About page doesn't name a funder.

Hendrycks and Oliver Zhang founded CAIS in 2022. Wikipedia gives the year only, not the month or day. The group is based in San Francisco, and its website is safe.ai. Its stated mission reads: "Our mission is to reduce societal-scale risks from artificial intelligence."

Its work spans technical AI safety research, AI ethics research, advocacy, and support for growing the AI safety research field. CAIS also runs a compute cluster for AI safety research. Its About page describes four teams: Research, Cloud and DevOps, Projects, and Operations.

What the Filings Show

The table below shows the total revenue, expenses and assets that the two entities reported to the Internal Revenue Service (IRS) for each fiscal year. Both entities' fiscal years end in December.

Entity	Fiscal year	Total revenue	Total expenses	Total assets
CAIS (501(c)(3))	2022	$6,660,060	$816,760	$5,943,414
CAIS (501(c)(3))	2023	$16,089,191	$8,134,449	$13,989,499
CAIS (501(c)(3))	2024	$10,235,085	$7,163,607	$12,638,754
CAIS Action Fund (501(c)(4))	2023	$718,870	$139,139	$607,985
CAIS Action Fund (501(c)(4))	2024	$3,577,153	$2,394,762	$1,890,527

Revenue ran above expenses in every year in the table. For fiscal 2024, the 501(c)(3) took in $10,235,085 and spent $7,163,607, a difference of $3,071,478. The Action Fund took in $3,577,153 and spent $2,394,762, a difference of $1,182,391.

These totals don't say who gave the money. Open Philanthropy, which appears to now operate under the name Coefficient Giving, is widely reported to be a major CAIS funder. No public source establishes whether CAIS takes money from AI companies.

Who Runs It

The About page lists Hendrycks as Executive and Research Director and Oliver Zhang as Managing Director. The FY2024 filing lists Hendrycks as Executive Director with reported pay of $314,534. Nicholas Beckstead, Policy Lead, reported $198,071, and Bri Treece, COO, reported $169,980.

Two unpaid directors appear in the annual Form 990 filings for fiscal 2022 to 2024. Rune Tybirk Kvist and Jaan Tallin, spelled that way in the filing, each report $0. The filings don't establish whether either one is a donor.

CAIS doesn't publish a headcount. The filing's list of eight people with reported pay is not a staff total.

Hendrycks has his own profile in our Dan Hendrycks explainer.

The 2023 Statement on AI Risk

In May 2023, CAIS published a one-sentence statement: "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war." The statement names extinction as the risk it covers.

CAIS's About page says more than 600 AI experts and public figures signed the statement. Wikipedia describes the signers as hundreds of AI professors, plus leaders of major AI companies and other public figures. The statement page shows no total, so the exact number isn't published.

The statement page names signers such as Geoffrey Hinton, Yoshua Bengio, Demis Hassabis, Sam Altman and Dario Amodei. Amodei and Altman run AI companies. Critics read their support as regulatory capture, because the heads of the companies the statement warns about endorsed it. Supporters read it as broad consensus.

For the wider debate over AI extinction claims, see whether AI is dangerous and our guide to the AI doomers.

What the Center for AI Safety Does

CAIS lists several programs on its About page, and Wikipedia adds its research output. The main ones are:

A compute cluster that, per the About page, "supports c. 20 research labs." More than 100 researchers have used its computing infrastructure since March 2023.
An online course called Intro to ML Safety. The About page says almost 300 students have received training through CAIS.
Events that the About page says more than 500 estimated machine learning (ML) researchers have taken part in.
A Trojan Detection competition that drew 1,200 submissions from over 70 teams.
A fellowship for philosophy professors to work on conceptual problems in AI safety.

CAIS researchers published "An Overview of Catastrophic AI Risks" in 2023. The authors were Hendrycks, Mazeika and Woodside, and the paper is arXiv:2306.12001. It covers autonomous warfare, engineered pandemics, and AI deception and hacking.

CAIS also worked with university researchers on an automated method for finding adversarial attacks on large language models (LLMs). Those attacks bypass safety measures. Wikipedia frames the work as showing that current safety systems are inadequate.

The Lobbying Arm and the Two-Entity Split

CAIS has two legally separate entities. The 501(c)(3) is Center For Artificial Intelligence Safety Inc. The 501(c)(4) is Center For Ai Safety Action Fund Inc. Both are based in San Francisco, California.

Pirate Wires describes the Action Fund as CAIS's lobbying arm. ProPublica's Nonprofit Explorer records that donations to the Action Fund are not tax deductible. It classifies the Action Fund as a civic league or social welfare organization.

The two entities are legally separate, and each has its own EIN. Dan Hendrycks is President of the Action Fund at $0 pay in both the FY2023 and FY2024 filings. At the 501(c)(3), he is paid as Executive Director.

On the Action Fund side, Varun Krovi is listed as Executive Director in FY2024, with reported pay of $274,168 plus $19,807 in other compensation. Nathan Calvin is listed as Senior Policy Counsel, with $122,049 plus $10,231. In FY2023, Krovi appears as "Director" at $80,208 plus $1,033.

The FY2023 filing also lists Bri Treece as Secretary, Kelsey Lawler as Treasurer until the term ended in December 2023, and Anthony Barrett as Treasurer from December 2023. All three are at $0. Treece is also the 501(c)(3)'s COO, at $169,980 in FY2024.

The Action Fund's revenue rose from $718,870 in FY2023 to $3,577,153 in FY2024. The filings don't show how much money moved between the two entities.

SB 1047 and the Criticism It Drew

California's Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, Senate Bill 1047 (SB 1047), was introduced on February 7, 2024. State Senator Scott Wiener authored it. The CAIS Action Fund sponsored the bill, and its listed sponsors were the Center for AI Safety, Economic Security California and Encode.

The bill would have applied to models trained with more than 10^26 operations that cost over $100 million. It required a pre-training safety protocol, a shutdown capability often called a "kill switch," and annual third-party audits from January 1, 2026. It also included whistleblower protections and a public compute cluster called CalCompute.

The bill fits into the wider policy debate covered in our AI regulation guide.

The Senate passed the bill 32-1 on May 21, 2024, and it was amended on August 15, 2024. The Assembly passed it 48-16 on August 28, 2024. The Senate re-voted 30-9 on August 29, 2024, and Governor Gavin Newsom vetoed it on September 29, 2024.

The bill drew heavy named opposition. Fei-Fei Li argued on August 6, 2024 that it would harm the US AI ecosystem. Andrew Ng called instead for more targeted measures, including work on deepfake pornography, watermarking of generated content and funding for red teaming.

Yann LeCun argued the bill would kill open-source AI models, and critics warned that liability could stop companies from releasing their models. Nancy Pelosi, Zoe Lofgren and Anna Eshoo were among the congressional opponents. Andreessen Horowitz opposed the bill and ran a website against it.

Industry groups, including California business and tech trade associations, also opposed it. Y Combinator also opposed it. Newsom's veto message said the bill did not take into account whether an AI system is deployed in high-risk environments. He warned it could "give the public a false sense of security."

A criticism of CAIS's own role comes from a Pirate Wires conflict-of-interest critique published in June or July 2024. It argued that the lobbying arm co-sponsored SB 1047 while CAIS's founder, Dan Hendrycks, had a stake in Gray Swan AI. Gray Swan is an AI auditing company, and the critique said its tools appeared positioned to serve the bill's auditing requirements.

Gray Swan responded that it does not hold any position on SB 1047. It said its products "are not and will not be designed to satisfy any of the auditing requirements proposed in the bill."

CAIS's own newsletter faced a disclosure complaint too. A reader's comment on the newsletter said it failed to disclose Hendrycks's adviser role at xAI when it wrote about Musk and SB 1047. CAIS replied that it "didn't bother since I assumed the ~all readers were aware."

For the other groups in this fight, see our who is fighting AI guide.

Is the Center for AI Safety Legit?

The Center For Artificial Intelligence Safety Inc is registered with the IRS as a 501(c)(3) nonprofit. Its employer identification number (EIN) is 88-1751310, and it is based in San Francisco, California. The IRS issued its tax exemption in January 2023, according to ProPublica's Nonprofit Explorer record.

The CAIS Action Fund is registered as Center For Ai Safety Action Fund Inc, with EIN 93-2442608. It is a 501(c)(4), and its tax exemption was issued in October 2023. Its donations are not tax deductible, according to its ProPublica record.

The named leaders are Dan Hendrycks, Executive and Research Director, and Oliver Zhang, Managing Director, as listed on the safe.ai About page on October 21, 2026. On the Action Fund side, the FY2024 filing names Varun Krovi as Executive Director and Nathan Calvin as Senior Policy Counsel.

The dated record runs as follows. CAIS published its statement in May 2023. SB 1047 was introduced on February 7, 2024, passed the Senate 32-1 on May 21, 2024, and was vetoed on September 29, 2024. Annual filings for fiscal 2022 to 2024 are on ProPublica.

Two gaps remain. State-level charity and lobbying registrations aren't checked here. CAIS doesn't publish a donor list, so there is no donor list to check. The criticism above concerns conduct and disclosure, not these registration records.

To see the registration details yourself, open the 501(c)(3) record on ProPublica and the Action Fund record. Both show the EIN, the exemption date and each year's filed totals.

Sources

Frequently asked questions

▸ Is Center for AI Safety legit?

The Center For Artificial Intelligence Safety Inc is registered with the IRS as a 501(c)(3) nonprofit, with EIN 88-1751310, based in San Francisco. The IRS issued its tax exemption in January 2023, according to ProPublica's Nonprofit Explorer. Dan Hendrycks is listed as Executive and Research Director on its website. Its fiscal 2024 filing shows $10,235,085 in total revenue.

▸ Who funds the Center for AI Safety?

The Center for AI Safety doesn't publish a donor list, and its About page doesn't name a funder. Its tax filings show total revenue and expenses, but not the names of donors. Open Philanthropy is widely reported to be a major funder, but no grant amounts could be confirmed.

▸ Who runs Center for AI Safety?

Dan Hendrycks is Executive and Research Director, and Oliver Zhang is Managing Director, according to the CAIS About page. Hendrycks and Zhang founded the nonprofit in 2022. Its fiscal 2024 filing lists Hendrycks as Executive Director with reported pay of $314,534.

▸ What is the center for AI safety?

The Center for AI Safety (CAIS) is a San Francisco nonprofit working to reduce societal-scale risks from artificial intelligence. It runs a compute cluster for AI safety research, an online course called Intro to ML Safety, and research on catastrophic AI risk. It also organized the 2023 Statement on AI Risk, which CAIS says more than 600 experts and public figures signed.

▸ What does the Center for AI Safety do?

CAIS says more than 100 researchers have used its computing infrastructure since March 2023, supporting about 20 research labs. It says almost 300 students have received training and over 500 estimated machine learning researchers have taken part in its events. It also ran a Trojan Detection competition that drew 1,200 submissions from over 70 teams.

▸ What is the CAIS Action Fund?

The CAIS Action Fund is a separate 501(c)(4) sister organization, with IRS tax exemption issued in October 2023. ProPublica records that donations to it are not tax deductible. The Action Fund sponsored California's SB 1047 AI safety bill, which Governor Gavin Newsom vetoed on September 29, 2024.

Latest related briefings

REGULATION POLICY

Anthropic’s $1B AI Risk Bill: What It Means for People Now

A proposed law would let the government block AI models costing over $1 billion. Here’s what this means for workers, families, and students today.

Oct 21, 20265 min read

Read analysis JOBS LABOR

Anthropic AI Workforce Study Flags Shifts for Workers

Anthropic's AI workforce study signals shifting job roles and security, raising concerns for workers and families as automation accelerates in 2026.

Oct 21, 20264 min read

Read analysis REGULATION POLICY

Student Walkouts Demand AI Regulation, Safer Schools

High school students are walking out to protest gun violence and demand AI regulation, showing growing concern for safety and technology’s impact on their futures.