cG made more AI policy grants than this in 2025! E.g. just from the >$1M bucket, there was also Oxford AI governance Initiative, Global Shield Global Catastrophic Risks Advocacy, Center for a New American Security, GovAI, Tarbell Center for AI Journalism, and maybe others I am missing!
Thanks Oscar!! Yeah searching "policy" was not the most robust methodology lol. Though worth noting, I was mainly focused on identifying c4 bets rather than c3 ones since political activity is far less restricted for c4 orgs. I think all of the orgs you mentioned are still c3s! If you know of any major c4 bets, lmk :)
Thank you for this, your rigorously clarified a lot of thoughts I required and I’m sure many needed.
Most of us at first though may think AI Safety goes into where our intentions believe it’ll end up going to, rarely ever the case. And in this domain where valuations are becoming scary to say the least.
Agree with the broad take. I'd noticed this from my own slightly different pov: I have 5-6 years of advocacy/lobbying experience, strong network in relevant government departments, and want to pivot into AI safety... and I can't for the life of me find any roles to apply for! I've seen one or two in the past ~6-12 months of looking (UK based)
(first ever comment here go easy on me)
Thank you for this post!
I was quietly thinking that, in the AI safety space, advocacy is more important than research at this point, precisely for the reasons you outline in the 'research vs lobbying section'. Thank you for giving a voice on the EA forum to these concerns of mine!
I would tentatively add that we already know that policy solutions (e.g. AI pause, global cooperation, mandating safety-first approaches to AI products) would be most effective in mitigating most X- and S- risks from AI (as opposed to finding technical solutions to the risks), but because of the general belief that policy is less tractable than technical improvement, we observe a lot more energy and funding being directed towards technical research into making AI products safer (which, as you say, isn't a sinecure because you can find technical solutions all you want, they won't necessarily be implemented at scale unless you have robust legislation making this compulsory!) rather than policy.
However, this is a self-fulfilling prophecy. The less people believe policy is tractable, the less they invest in it, and the less tractable it actually is.
I'm interested in the reasons this happens and I can think of two possible ones:
I have actually made some recommendations on how this could be mitigated which you can read about here. Separately, because of all the recently made redundant policy specialist at e.g. USAID, there's probably an opportunity to reverse that gap quite quickly if the movement becomes more receptive to policy/lobbying as a priority.
This was good, thank you! I didn't realize some of these orgs existed. Did you see the recent post where ControlAI is asking for $50 million?
Thank you so much, Jenna! I hadn't seen that, have updated the post to include this information!
Executive summary: The author argues that donors who want to maximize AI safety impact should redirect giving from well-funded research nonprofits (c3s) toward underfunded political advocacy (c4s), even at the cost of losing tax deductions.
Key points:
This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, and contact us if you have feedback.
*if you want us to win; choose advocacy instead
In one sense, AI safety is not funding-constrained. There are hundreds of millions of dollars flowing into research orgs, more coming with the Anthropic IPO, and grantmakers are already struggling to deploy what they have.
And yet: the AI industry spent over $100 million lobbying the federal government in 2025, while the biggest AI safety c4 spent $310,000.
The highest-leverage path for donors looking to give to AI safety right now is to give less money, but to more neglected destinations.
Specifically, you should strongly consider routing what would have been a c3 gift to a c4 instead, calibrated to equivalent personal cost.
501(c)3 organizations are very limited in terms of the amount of money and time they can spend on lobbying and political advocacy-shaped work. They’re also strictly prohibited from participating in political campaigns or supporting candidates:
Under the Internal Revenue Code, all section 501(c)(3) organizations are absolutely prohibited from directly or indirectly participating in, or intervening in, any political campaign on behalf of (or in opposition to) any candidate for elective public office… Violating this prohibition may result in denial or revocation of tax-exempt status and the imposition of certain excise taxes.
Donations to 501(c)(3)s are tax-deductible, but the trade-off is that these organizations can’t do much by way of direct political advocacy— unless they’re willing to put their c3 status at risk of being revoked entirely.
In contrast, 501(c)(4)s can spend unlimited amounts on lobbying and engage in direct political advocacy, as long as their primary purpose remains social welfare. Donations to c4s aren’t tax-deductible, but they have far more freedom to engage in the political process.
AI safety 501(c)(3) work is substantially less funding-constrained than 501(c)(4) work. As a result, the marginal dollar spent on lobbying and political advocacy is more important than the marginal dollar spent on research.
Here are some examples of grants given to organizations in the AI safety space last year:
All of the above are 501(c)3 organizations.
To be clear, these are amazing orgs! This is only meant to highlight the amount being given to c3s-- tens of millions.
Let’s zoom in on CG for a moment, since it’s the biggest funder in the field. Their TAI fund appears to have made 158 grants in 2025:[2]
Of the 158 listed, a grand total of three appear to have been for policy work:
Notably, none of the above organizations can engage in substantial lobbying or political campaign work. RAND is a c3, IAPS is fiscally sponsored by a c3, and Training for Good received funding earmarked for a policy fellowship.
In researching CG’s grant database, I did find one notable c4 bet: $3M in general support to Americans for Responsible Innovation in August 2024. ARI is bipartisan, which may explain why it cleared CG's reputational bar (more on this below and in my original post about the IPO). It’s a good start, but it appears to be an exception rather than a pattern. Benjamin Todd’s January 2025 post on 80,000 Hours noted that CG had recently stopped funding “many Republican-leaning think tanks, such as the Foundation for American Innovation,” and I haven't been able to determine whether their support for ARI has continued.
(I encourage you to spend some time on Coefficient Giving’s Navigating Transformative AI Fund grant database and examine the grants yourself to verify the c3/c4 asymmetry.)
So a bunch of c3 organizations in the AI safety space received tens of millions of dollars each last year. Meanwhile, the AI industry was deploying its own dollars in a very different way:
Eight of the largest tech, AI, and social media companies spent a combined $36 million on federal lobbying during the first half of 2025 — an average of roughly $320,000 per day that Congress has been in session.
Right now, the industry is winning the legislative war because the safety side is not putting up a fight.
Technical safety research only matters if frontier labs actually implement it. Even if Anthropic adopts the latest alignment techniques voluntarily, other companies probably won’t if it slows them down or cuts into margins.
Without legislation, every safety advance is effectively optional— a recommendation companies can take or leave depending on competitive pressure.
Legislation is what makes technical safety advances mandatory for frontier labs. It’s also what buys time. Every regulation that imposes a meaningful evaluation requirement, every disclosure mandate, every liability framework, every whistleblower protection policy slows the race and gives safety researchers more runway to solve the problems we don’t yet know how to solve.
While we have tens of millions of dollars pouring into major organizations in the c3 space, we don’t have nearly enough funding going towards political AI safety work. The gap between funding for research and funding for advocacy is enormous.
As I wrote about previously, part of this is due to the concentration problem: currently, the vast majority of AI safety funding flows through a handful of institutional funders. Some of those funders (including the vast majority of private foundations,[3] most DAFs, and corporate matching programs) are structurally barred from significant C4 giving.
Others, including CG, can technically fund C4s but face reputational constraints that make them cautious about politically charged advocacy, especially work that opposes specific industry players. As a result, the entire field has inherited a bias toward C3 research over C4 advocacy.
Individual donors face neither constraint. This means the people who can actually fix this gap are individual donors, who are the only actors in the ecosystem without structural or reputational barriers to C4 giving.
A top-earning donor giving $1M to a c3 actually spends ~$650K after the federal deduction (capped at 35% under the One Big Beautiful Bill Act, effective 2026).
This means that $650K, given directly to a c4, represents an equivalent personal cost with arguably greater counterfactual impact.
The Formula
C4-equivalent = X − [(X − floor) × effective deduction rate]
Where:
So, for a donor with $5M AGI giving $1M to a c3, these are roughly the numbers:
Starting in 2026, the first 0.5% of a donor’s AGI in charitable giving confers zero federal tax benefit under the One Big Beautiful Bill Act’s new deduction floor.
For a donor whose total c3 giving falls at or below that floor, c3 and c4 donations will cost exactly the same, because the amount you’re actually donating is equal to your personal cost (there is no deduction). In this instance, the tax case for c3 donations disappears entirely,[4] and c4 wins on any positive impact multiplier.
As an example, let’s say we have an Anthropic engineer with $3M AGI who gives $15K to AI safety. The floor is $15K, so zero of that engineer’s $15K gift is deductible, meaning c3 and c4 are tax-identical.
I am not your financial advisor and would recommend discussing your specific case with a qualified CPA. For example, the above math doesn’t apply if you’re planning on donating stocks directly, because of capital gains tax (it’s complicated).
The orgs below take different approaches. Some focus on catastrophic risk directly, others on adjacent concerns like child safety, worker protection, or chip security. Read the descriptions and pick what fits your priorities.
Here’s the reality…
Congress is barely in session.
Big AI is moving at full speed.
And the window to put real safeguards in place is closing fast.—Humans First, D.C. Town Hall Tour
Center for AI Safety Action Fund
Read more: https://action.safe.ai/work
Some practitioners in the AI safety policy space believe donations to politicians and PACs are meaningfully more impactful than the best C4s. Zach Stein-Perlman, who advises donors in this area, has estimated roughly “5x as impactful.”
I haven’t independently verified this, but I think it could be true, depending on assumptions surrounding timing, electoral leverage, and the specific donation opportunities Zach might point you towards. (Also worth flagging: Meta is spending $65M this year on super PACs designed to elect AI-friendly state officials— the biggest election investment in the company's history. That’s a pretty good sign this kind of giving works.)
However, there are trade-offs: PAC contributions are publicly disclosed, more legally complicated, require US permanent residency, and there are contribution limits; there might also be reputational risk for recipient political candidates depending on who you are (e.g. see Carrick Flynn case study; I think being backed by SBF almost certainly hurt his odds).
That said, if you’re considering donating $20K+ and willing to handle the complexity and public disclosure, PACs may be worth exploring. Zach has offered to advise on this (more on how to reach him below).
For most readers, the C4 case I’ve made above is probably still the right starting point.
The AI safety field is flush with money, yet starving for advocacy. The next billion dollars to CG or a DAF probably won’t solve this problem.
If you care about AI safety and you were planning to donate this year: before defaulting to a c3, run the math above and seriously consider whether a c4 gift at equivalent personal cost would do more.
If you want to talk this through before giving, here are some options:
Pick whoever fits. Different people are better positioned for different donor profiles.
This is an opportunity to maximize your counterfactual impact in a way that almost no other giving can. Giving well is hard. It takes research, judgment, and time. It would be easier to park money in a DAF and forget about it, or forward donations to whoever’s most legible, or give to people you like, or just not give at all.
But the political work that needs funding right now is work that most institutional funders can’t or won’t touch. The dollars sitting in your account are structurally different from every other dollar in the ecosystem, and deploying them well might matter more than any other decision you make this year.
Thank you to Jack Douglass for helpful discussion & feedback.
Note: I’m mostly using c3 vs c4 as a shorthand for research vs. advocacy; it’s important to note that some c3s do engage in lobbying (though, again, they are legally limited in how much they can do). Also, there are other ways to donate to advocacy-shaped things, such as donating directly to PACs and political campaigns; I talk a bit more about direct donations of these kinds later on.
Caveat: their page says “featured grants,” so it’s possible that there are more grants not listed on the page, but I don’t know how to verify this. The asymmetry with respect to lobbying and advocacy still pretty clearly exists, though (keep reading for more stats!)
Foundations can technically donate to c4s, but only for non-partisan activities. Even then, “private foundations must follow a specific grant-making process called expenditure responsibility when providing grant funds to 501(c)(4)s because they are not public charities. For this reason, private foundations have historically preferred to limit their grantmaking to 501(c)(3) public charities to avoid navigating these more onerous rules” (Alliance For Justice).
(State-level deductions may still apply depending on where you live— another reason to talk to a CPA about your specific situation.)
Flagging again: this donation is less than a sixth of Leading the Future, and the money “isn’t allowed to be used in the midterm battles.”
I haven't personally chatted with Zach Stein-Perlman. I plan to do so as I do deeper research into PACs + other c4 orgs not on this list, for a separate project. I'm flagging his public offer to advise donors, not necessarily endorsing the specific advice he'll give— if you reach out to him, you may get framing that diverges from this post, and either could be right for your situation.
Thanks for this post! I broadly agree with your introduction but I'd say the "research vs. advocacy" distinction is more important than "501(c)(3) vs. 501(c)(4)". There are 501(c)(3) advocacy orgs including MIRI, Palisade, PauseAI US/PauseAI Global, Lightcone Infrastructure (sort of), etc. Lobbying on policy measures is one component of advocacy, but not the only component. I'm uncertain about how valuable it is compared to other kinds of advocacy that 501(c)(3)s can do.
Hi Michael! Thanks for the comment. Agreed that research vs. advocacy matters more than the c3/c4 distinction; I have a footnote clarifying this, but in retrospect it probably belongs in the body of the post.
That said, I think c3/c4 is a useful shorthand for donors specifically because it maps onto who else will fund this work. c3 advocacy orgs like the ones you mention (MIRI, Palisade, PauseAI, maybe even Lightcone but uncertain / don't know enough about their work beyond Lighthaven lol) can receive funding from private foundations, DAFs, and corporate matching programs (whether or not they do in practice is a different question, and I don't know enough about that to comment, but my sense is that broadly, c3s are way less funding constrained than c4s for structural reasons as c4 lobbying orgs largely can't receive funding from these kinds of sources at all).
So while the intellectual distinction is research vs. advocacy, I'd say the funding gap is sharpest at the c3/c4 boundary, and that's where an individual donor's marginal dollar is most counterfactually valuable-- other than perhaps donations to PACs/specific politicians.
I'm also uncertain about the relative value of lobbying vs. other forms of advocacy. My intuition is that direct lobbying is underfunded relative to its impact, partly because the structural barriers to funding it are so high, but I'd love to see more analysis on this.
All the c3 orgs I mentioned are funding-constrained (except perhaps MIRI). AI x-risk advocacy is unpopular among big funders (with a couple exceptions like SFF), which means both the c3s and c4s are funding-constrained. At least that's true currently; I'm not sure if that will still be true a year from now.
I agree with you about the structural funding disadvantage to c4s, but empirically it doesn't look obviously important for driving funding constraints.
Yeah this is a difficult question, I don't know. Another problem is that lobbying is opaque so it's harder to tell who's doing a good job. (ControlAI is my favorite lobbying org because they write a lot about what they do and I like what they write. But also I reached out to them and they said they were
not accepting donationsedit: only accepting donations of $100K or more.)Thanks Michael! Will be doing more research into the c3s you listed, as I'm working on compiling a complete set of funding recommendations for a future post -- if you have recs on things to read, would be useful :)
Re: ControlAI, seems that they're currently fundraising, but only accepting gifts of $100k+. Have updated my post to reflect this.
Since I'm biased on the matter, I will start by linking the posts I've written:
Eric Neyman and Zach Stein-Perlman write recommendations on AI risk advocacy, most of their work is non-public but Eric wrote Consider donating to Alex Bores, author of the RAISE Act.
Zvi wrote The Big Nonprofits Post [2024] and 2025. He's a grantmaker for SFF which is my favorite of the big grantmakers.
Coefficient Giving and Longview Philanthropy often write about their grantmaking decisions (although not in as much detail as I'd like). I tend to have some pretty big disagreements with them, but they're still worth reading.
Most reasoning on grantmaking/donations happens in private, so there's not a whole lot to read. If you broaden the question to writings on general strategy (not just donations), there's a ton of stuff worth reading. I will just link two that align best with my personal views:
I missed that, thanks for calling it out!