Written by Eric Neyman, in my personal capacity. The views expressed here are my own. Thanks to Zach Stein-Perlman, Jesse Richardson, and many others for comments.

Over the last several years, I’ve written a bunch of posts about politics and political donations. In this post, I’ll tell you about one of the best donation opportunities that I’ve ever encountered: donating to Alex Bores, who announced his campaign for Congress today.

If you’re potentially interested in donating to Bores, my suggestion would be to:

1. Read this post to understand the case for donating to Alex Bores.
2. Understand that political donations are a matter of public record, and that this may have career implications. Decide if you are willing to donate to Alex Bores anyway.
3. If you would like to donate to Alex Bores: donations today, Monday, Oct 20th, are especially valuable. You can donate at this link.

Or if you’re just curious, read whatever parts of the post you find most interesting!

Introduction

In June, Zvi Mowshowitz wrote a post about the New York RAISE Act. I'd encourage you to read Zvi’s whole post (and I'll include a brief summary below), but Zvi's bottom line was:

The bill is insufficient on its own but an important improvement upon the status quo. I strongly support this bill.

The RAISE Act is one of only a handful of bills that are specifically aimed at mitigating catastrophic and existential risks from AI. (The bill passed the New York legislature, and will likely be signed by the governor in the coming months.^[1])

The bill's sponsor is a state legislator named Alex Bores (a Democrat). After seeing California SB 1047 get vetoed, Bores decided to make AI safety his first priority. He sponsored the RAISE Act and fought tooth and nail to get it through the legislature, spending a huge amount of political capital along the way (see below).

Today (Monday, October 20), Alex Bores announced that he’s running for the U.S. House of Representatives. Quoting the New York Times:

Mr. Bores, a former software engineer, said the rapid pace of technological advancement and its implications for American democracy compelled him to run. He pointed to President Trump’s close relationships with wealthy tech executives as well as the omnipresence of artificial intelligence and other software that can distort reality, saying he felt that most Democratic leaders were ill-equipped to contend with such challenges.

Although he isn’t considered the favorite, I think he has a decent chance of winning. For reasons that I go into below, I’m really excited about his run for Congress, and think that electing Bores would be really beneficial for having sensible AI regulation in the United States. I also think that giving to Bores’ campaign is one of the best opportunities I’ve seen so far for mitigating existential risk from AI. I plan to donate $7,000 to his campaign (the legal maximum), and I think many readers of this post should too.

Note that giving to Bores on Monday, October 20th, is significantly more valuable than giving later. This is because campaigns often come out with a press release announcing how much they raised on day 1. This helps the candidate secure endorsements and sometimes convinces other potential candidates to stay out of the race. (See more details here.)

(If you’re hearing about Bores for the first time today, October 20th, and are feeling rushed to make a donation: yup, I’m sorry about that. See this section of the post for some thoughts about that.)

If you’re already convinced to donate to Alex Bores, you can donate here (and thank you so much!). My donation suggestion is $7,000 (the maximum), if you can afford it. (But note: before deciding to donate, please be aware that political donations are a matter of public record! If you’re potentially interested in working in the federal government, this could have career implications for you. I discuss those implications below.)

If you want to know more about the case for donating, read the rest of this post 🙂

(Note: my goal is to provide an unbiased best-guess assessment of this donation opportunity, and I’m worried about people being more reluctant to share arguments against donating than arguments in favor of donating. If you think there are substantial downsides to donating that I haven’t touched on, please leave a comment, or email me.)

Before doing an explicit cost-effectiveness analysis, I want to mention some things that I particularly like about Alex Bores.

Things I like about Alex Bores

The most important thing I like about Alex Bores is that he has a great track record on AI safety. Concretely:

• He sponsored the RAISE Act. In a nutshell, the RAISE act requires large model developers(ones with >$100 million training runs) to:
  • Report “safety incidents” to the attorney general and Division of Homeland Security and Emergency Services. (This is similar to how companies must report cybersecurity breaches.)
  • Have and follow a safety plan to prevent severe risks.
  • Not release models that would create an unreasonable risk of “critical harm” (roughly speaking: the death or serious injury of one hundred or more people, or at least one billion dollars of damages).
• He fought tooth and nail to get the RAISE Act through the legislature, spending a huge amount of political capital along the way.
  • In the New York legislature, not all bills get voted on. A bill only gets voted on if the NY Senate majority leader and the Speaker of the NY House decide to put it up for a vote. In this case, putting the bill up for a vote was politically costly: that's because the lobbying arms of big-tech corporations were spending huge amounts of money to sink the bill, with an implicit threat to spend money against legislators who would help the bill pass. Ordinarily, such opposition would have doomed the bill, even though the bill had overwhelming support. However, particularly effective legislators (and Alex Bores is one of them^[2]) can build up enough political capital with the legislative leadership that they can occasionally spend that capital on bills that they particularly care about. Bores did that with the RAISE Act.
• He understood the ins and outs of the bill really well, as demonstrated in his handling of adversarial questions on the House floor and his appearance on the Cognitive Revolution podcast.
  • Note that this is relatively uncommon: most legislators don’t understand the AI bills that they sponsor particularly well.
• I have talked to several AI policy experts who have worked with Bores. They have been consistently impressed with his understanding of the technical details of AI governance, and his dedication to making sure that advanced AI is developed safely.

While this is the most important consideration for me, there are other things I like about him, too:

• I recently met him, and he struck me as an unusually honest person. I’ve chatted with several high-profile politicians, and while none of them were dishonest with me, they sometimes tried to modulate their language to try to please me. I didn’t get that sense at all from Alex. He told me exactly what he thought on every issue I asked him about, in a way that I really appreciated.
• One of his top priorities (in addition to AI safety) is safeguarding American democracy and preventing democratic backsliding. He plans to do that by trying to pass legislation that limits the power of the presidency. I’m really glad that he plans to prioritize this.
• He cares about animal welfare. For example, he led a budget letter^[3] to establish an alternative protein research center in NY. He is a vegetarian.
• He co-sponsored a bill to institute a pilot program for implementing a land value tax in New York. He also just introduced a bill to prevent sportsbooks from limiting winning players (which I consider to be a good idea for reasons described here). This makes me think that Bores is more generally a technocrat who will support important, carefully considered policies.
• He has a masters degree in computer science. (In fact, he is the first New York state legislator to have any degree in computer science!) I think that his technical background is pretty useful for evaluating tech regulation bills.

Are there any things about Bores that give me pause?

Alex Bores worked at Palantir for a few years. I think that many of Palantir’s activities are pretty evil. That said, as far as I know, Bores did not work on any such projects while at Palantir, and he says that he left Palantir over concerns about its collaboration with ICE.

Bores also highlights his ties to unions and support for organized labor. I generally have mixed feelings about unions, and support for organized labor is often a sign that a politician is pretty left-wing (whereas I’m center-left). That said, my overall impression is that Bores is center-left, rather than left-wing.

So: maybe some really minor things? But, pretty much no.

Cost-effectiveness analysis

This section is broken down into two subsections:

1. How does an extra $1k affect Alex Bores’ chances of winning his race?
2. How good is it if Alex Bores wins his race?

My estimate of #1 will be much more precise, because it’s a pretty straightforward statistical modeling question, but the two parts are equally important to the analysis.

How does an extra $1k affect Alex Bores’ chances of winning?

My all-things-considered guess is that Alex Bores has a 20% chance of winning. My best guess is that a marginal $85,000 donated on Monday, October 20th raises his chances of winning by 1%. (My 50% confidence interval is something like [$40k, $170k].) I also estimate that a marginal $105,000 donated sometime in 2025 but after October 20th raises Bores’ chances of winning by 1%. This breaks down as follows:

• I estimate a 1.6% chance that Bores loses his primary election, but by fewer than 1000 votes.
• I estimate that the Bores campaign could spend a marginal $300 on advertising in order to attract the support of one more voter.^[4] (Note: low confidence; I wouldn’t be surprised if this is off by a factor of 2 in either direction.)
• I estimate that money given in 2025 is 2x more valuable than money given in 2026, because showing strength in early fundraising numbers consolidates party support (or prevents opponents from consolidating support), and that money given on Monday, October 20th is 2.5x more valuable than money given in 2026.
  • Erik Bottcher, one of Alex Bores’ primary opponents, raised $683k on his first day (a New York state record), which generated positive headlines and established him as a serious candidate. The buzz around Alex’s candidacy announcement will be qualitatively different if he can beat Bottcher’s number, especially if he beats it by a lot. I would guess that Bores will beat Bottcher’s number, but I don’t think it’s a slam-dunk.
• I adjust my cost-effectiveness estimate by 10% to take into account the fact that the big tech super PAC might get involved (or spend more) if Bores raises a lot of money.

Multiplying these numbers together gives the estimate above. I’m not sure how interested readers will be in the details, so I’ve put my reasoning in an appendix.

How good is it if Alex Bores wins?

So, suppose that you believe my estimate that a marginal $85,000 given on Alex Bores’ launch day increases his chances of winning by 1%. How effective is such a donation?

We need to think about how much impact we might expect Bores to have, compared to alternative ways that we could use $8.5 million.^[5] To do that, I’ll start by listing what I consider to be the most important outcomes from Bores winning this race.

Direct influence on legislation

I see three main ways that Alex Bores could come to directly influence legislation in a way that benefits AI safety.

The first of these is influencing big spending bills on the margin, to direct more government money toward AI safety. Every year, the government passes a $7 trillion budget; in many years, it additionally passes one other multi-trillion-dollar spending bill. These bills are big enough that individual representatives can often influence how a small fraction (but still huge absolute amount) of the money is spent. A representative whose first priority is AI safety could plausibly cause a line item allocating $1 billion to AI safety to be included in the budget.

The second of these is influencing AI legislation: either by including safety-related provisions in sweeping AI legislation, or by making safety-specific legislation more sensible/targeted at catastrophic risks. If Bores wins, he might be the only member of Congress who chooses to prioritize AI safety, and who has any amount of expertise on the matter. This definitely doesn’t mean that he’ll get his way on AI safety issues, but I think that he will be listened to, and will have some influence.

The third is sponsoring major AI safety legislation. While we are not currently in a political environment in which major AI safety legislation can pass, this could easily change, either because people become more scared of AI, or because we get a presidential administration that is more favorable to AI regulation (e.g. in 2029). If that happens, Alex Bores could potentially sponsor and advocate for a major piece of AI safety legislation (analogous to California’s SB 1047) at the federal level. Bores says that he will pursue AI safety policies federally, if elected.

The House is a first step toward even more influential positions

Alex Bores is running for a seat in the House of Representatives, but many House members later go on to serve in even higher-ranking roles.

I’d guess that the conditional probabilities of Bores eventually taking these roles (if he gets elected to the House) are something like:

• Senator: 4%
• Cabinet-level official: 4%
• Governor: 2%
• President: 0.2%

I think this matters a lot. If Bores becomes a cabinet-level official (e.g. Secretary of Commerce), it’ll likely be a role where he would have considerable influence over the presidential administration’s AI policy.

Encouraging more action in this space

OpenAI president Greg Brockman and a16z co-founder Marc Andreessen recently created a Super PAC to fight against AI regulation, with a $100 million war chest. I expect that big tech will spend millions of dollars against Alex Bores in the primary election. If Bores can demonstrate that it’s possible to win his race even while standing up to big tech, I think that would encourage other legislators to sponsor bills like SB 1047 and the RAISE Act, despite opposition from big tech.

I also think that Bores’ election would encourage more people in the AI safety space to get involved in politics, either by contributing to efforts to promote AI safety via political giving, or by running for office.

How does this compare to other AI safety donation opportunities?

I’ll break other AI safety donation opportunities into three categories: technical AI safety; non-politics AI governance; and politics.

Comparison to technical AI safety

I think donating to Alex Bores compares very favorably to all technical AI safety opportunities that I’m aware of. I believe this for two reasons:

• The billionaires have it covered. I think a lot of the low-hanging fruit of funding the best technical AI safety agendas has been picked.
• I think the expected amount of counterfactual funding for AI x-risk focused technical safety work is quite large: on the order of $50 million/year if Bores is elected (see “Influencing big spending bills” above). I think these grants won’t be as good as the marginal technical AI safety grants funded by members of this community, but would still be pretty good. And that’s just one of several ways in which electing Bores would be beneficial to AI safety.

Comparison to non-politics AI governance

I think that one of the best interventions in AI governance is skilling up people who are interested in working on AI safety in the federal government (e.g. as a Congressional staffer).

I don't currently know of great ways to fund such efforts, but I think they are instructive as a comparison point. My admittedly wild guess is that it's possible to create a counterfactual career in x-risk-focused AI governance for about $500k-$1M. (This is assuming that a low two-digit percentage of their fellows counterfactually end up with such a career because of their fellowships.) So $8.5 million could be spent to create about ten AI governance careers. These are marginal AI governance careers, rather than the best ones; I’d guess that ten such careers are probably substantially worse in expectation than having Alex Bores in Congress (though maybe less than by a factor of 10).

Comparison to other political opportunities

Unfortunately there’s only so much I can say here, because most of those opportunities are sensitive. However, I believe that this is the best currently-existing political donation opportunity, by a factor of 2-3 or so.

I think the next-best donation opportunities (after Bores) are ones aimed at relationship-building with existing members of Congress. There are also a couple of other people running for office right now who are plausibly worth donating to, but I think those donation opportunities are worse by a factor of 2-10 or so.

I expect that big opportunities that are as exciting as this one will come up very rarely (maybe once every couple of years). One comparison point: it has been reported that Scott Wiener will run for Congress. I’m super excited about this, and think that we’ll only get big donation opportunities this good once a year or so. Nevertheless, I think that donating to Bores looks a little better, mostly because it’s clearer to me that Bores will continue to prioritize AI safety.^[6]

If you’re looking for the best way to spend your charitable donation budget this year, I’m not aware of a better opportunity than this one, and I don’t think that one will come up.

Comparison to non-AI safety opportunities

I think that electing Alex Bores to Congress would decrease x-risk by 1 in 5 thousand or so. Of course, this is sensitive to my particular views on AI, but I’ll spell out my calculations so that you can plug in your own numbers.

• The US government’s decisions in the lead-up to AGI will have huge effects on the trajectory of AI. The difference in my subjective probability of x-risk depending on whether the US government’s decisions on AI are generally wise or generally unwise is about 5%.
• Congress is a big part of that: its decisions don’t matter as much as those of the executive branch, but I think its actions are maybe 10% of the actions of the US government as a whole, weighted by importance.
• Having Bores in Congress would plausibly make Congress 2% better on AI, via some of the pathways discussed above.
• Plus, I think a lot of Bores’ expected impact (maybe half) would happen through routes other than being in Congress, e.g. being a cabinet secretary. (Thus, 2x multiplier.)

Multiplying these numbers together gives 1 in 5 thousand: in other words, electing Alex Bores saves about 1.6 million currently-existing people’s lives, in expectation; increasing his chances of winning by 1% saves 16,000 expected lives. Given my estimate that $85,000 donated on Monday, October 20th, increases Bores’ chances of winning by 1%, I think this donation opportunity compares very favorably with non-AI safety opportunities, even without taking into account future lives.

Logistics and details of donating

Who can donate?

Any US Citizen or permanent resident (e.g. any green card holder^[7]) can donate.

How much can I donate?

You can give up to $7,000: $3,500 for the primary election and $3,500 for the general election. If Alex Bores loses the primary election, you will get back any money you donate beyond $3,500.

Note that even though Bores is guaranteed to win the general election conditioned on winning the primary, those second $3,500 are still useful to him, for a couple of reasons. The first is that (as I discussed above) a large fraction of the value of a donation is that it helps Bores signal strength and consolidate support. The second is that representatives who raise more money tend to get better committee assignments, and it seems really useful for Bores to be on e.g. the Committee on Science, Space, and Technology.

In fact, since (by my estimate) Bores is only 20% likely to win the primary, there is an 80% chance that you will get back any money you donate beyond the first $3,500. This means that with probability 80%, the impact of those dollars is “free” (at least if you ignore time preference of money). The impact of the dollars is smaller, but above I estimated that 50-60% of the value of a donation is signaling campaign strength.^[8] For this reason, I believe that dollars beyond the first $3,500 are over 2x more valuable in expectation. If you’re donating $3,500 and can afford to donate $7,000, I think it’s likely that you should.

(Note also that this logic still applies if you think that e.g. only 30% of the value of a donation is in signaling campaign strength.)

How do I donate?

You can donate through ActBlue at this link!

If you are donating on October 20th: donations via ActBlue are preferred (credit card, Paypal, Venmo, Google Pay). While ActBlue donations are instant, checks and wire transfers made directly to the campaign’s bank account may take more than 24 hours to receive.

If you are donating after October 20th: feel free to donate via ActBlue, or consider doing a bank transfer. That’s because ActBlue (the vendor that donations go through) charges a 4% fee: if you donate $7,000, the Bores campaign will receive $280 less than that. I don’t currently know how to donate via bank transfer, but email me if you’re interested.

Will my donation be public? What are the career implications of donating?

Yes: donations are a matter of public record. You should take this into account when making your decision. The current administration has been illegally using ideological/loyalty litmus tests for hiring for some roles (e.g. rejecting people who say that Biden won the 2020 election). My best guess is that they are also discriminating on the basis of political donations, at least for some roles. So if you think you might want a role in the federal government, and you don’t already have a recent history of donating to Democrats, you should take this into account.

How big are the risks here? My current understanding (which could be wrong) is that:

• If you already have a big or recent donation history to Democrats, you have already paid most of the cost.
• If you don’t have a big or recent donation history to Democrats, are strongly considering a role in the federal government in the next few years, have a realistic shot at getting that role, and think that you would be much better than whoever they would hire instead of you, then you should probably refrain from donating.

For most AI safety or (especially) AI policy researchers who haven’t donated to Democrats in the last few years, small donations (e.g. $250) are not worth the cost in career capital. So if you work on AI safety, only donate if you’re donating a lot, or if you already have recent donations to Democrats on record.

Is donating worth the career capital costs in your case?

If you think it’s a close call, here are a few ways to try to figure out what you should do:

1. Check out the “donor profiles” section below, which gives some examples of potential donor profiles, and my subjective opinion on whether or not it makes sense for them to donate.
2. Check out this post, which has a questionnaire with items like “Do you have a security clearance? +4” and “Have you worked for at least 2 years in a big AI lab? +1". If your total score is at least 11, the post's recommendation is to avoid donating. (Note: I haven’t vetted the post super closely, but overall it seems reasonable to me.)
3. Check out the “more quantitative cost-benefit analysis” section below. (I think most readers of this post should skip this one.)

Some examples of potential donor profiles

In this section, I’ll give some hypothetical examples of potential donors to Alex Bores, and give my subjective assessment about whether – from an altruistic standpoint – it would make sense for them to donate $7k to Bores, or to forgo a donation to preserve career capital. (Note: these are my best guesses, and are subject to revision.)

(I’ve bolded “from an altruistic standpoint”, because there are of course personal costs to not being able to get a job. I’ll leave it up to readers of this post to price that in as appropriate.)

Technical AI safety researchers

• Example profile 1a: Alice is a technical AI safety researcher at an AI company or AI safety nonprofit. She isn’t the head of a safety team. She thinks it isn’t all that likely that she’ll end up in government, but thinks there’s some chance it could happen (e.g. if the lab is nationalized). She hasn’t donated to Democrats before.
  • My guess is that it would be altruistically optimal for Alice to donate to Alex Bores and similar opportunities.
• Example profile 1b: same as #1a, except Alice is considering quitting her current job and moving to Washington, D.C. to work on AI policy.
  • If Alice is only vaguely considering this option: probably she should donate.
  • If Alice is strongly considering this option, and is specifically looking at potential jobs in the federal government: probably she should not donate.
• Example profile 1c: same as #1a, except Alice is the head of an important safety team.
  • I think this case is borderline and probably depends on Alice’s particular circumstances.
• Example profile 1d: same as #1c, except that Alice made a $1,000 donation to a Democratic candidate for Congress in 2022.
  • I think Alice should probably donate: my guess is that from Alice’s 2022 donation means that she has already paid a substantial fraction of the career-capital cost of being a fairly consistent Democratic donor.

AI policy researchers

• Example profile 2a: Bob is a junior AI policy researcher in D.C. He’s a couple of years out of college, and just got a job at a nonpartisan think tank, where he researches AI policy. He thinks that AI policy is a good fit for his skills, and expects to continue working on AI policy. He isn’t specifically thinking about roles in the federal government, but thinks there’s a real possibility that he’ll want to work in the federal government in the next few years. Bob isn’t a strong partisan, and hasn’t donated to political candidates before.
  • I think Bob should probably not donate.
• Example profile 2b: same as #2a, except Bob strongly objects to the Trump administration’s actions, and would find it challenging to work in an environment where lots of people are in favor of the actions that Bob strongly opposes.
  • I think Bob probably should donate to Alex Bores. That’s because it’s unlikely that Bob would be a good fit for a policy role in the federal government under Trump.
• Example profile 2c: same as #2a, except Bob has some small, recent donations to Democrats on record.
  • I think this case is borderline, as Bob has paid some but not all of the cost of donating.
• Example profile 2d: same as #2c (recent small donations to Democrats), but Bob already has several years of experience and is pretty senior at his think tank and sees a realistic prospect of applying to work in the federal government in the next few years.
  • I think Bob should not donate.

A more quantitative cost-benefit analysis

This cost-benefit analysis below will let you decide two parameters:

• p, the probability that you will want to try to get a job in the federal government.
• r, the benefit to the world of having you in government, as a fraction of the benefit of Alex Bores being elected. (I realize that estimating r may be aversive to some readers; if so, I recommend that you base your decision off the “donor profiles” above instead, or on this post.)

Let’s say that you don’t have any recent donations to Democrats. Then I think that from an altruistic standpoint, it’s worth donating $7,000 to Bores (and donating to similar opportunities in the future) if:

• p * r < 2%, for technical AI safety researchers
• p * r < 1%, for AI governance researchers

The lower threshold for governance researchers reflects the fact that I expect more discrimination on the basis of donations for policy positions than for technical work.

(If you’re a smaller donor, the threshold should scale proportionally: e.g. an AI safety researcher donating half as much should only donate if p * r < 1%.)

My guess is that for a typical technical AI safety researcher, p is around 5-10% and r is around 1-3%. Assuming the numbers on the higher end, we still only get p * r = 0.3%. (And so, from an altruistic standpoint, I think that donating to Alex Bores is probably the right call for most technical AI safety researchers.) Meanwhile, for AI governance researchers, I think it will generally be a closer call.

I’ve put the details of my calculation in this appendix. I suggest that you read my calculation and see if it makes sense. I’m really uncertain of it, and I would really hate for my analysis to have adverse consequences because I made bad assumptions.

Potential concerns

What if Bores loses?

As I mentioned above, Bores isn’t currently considered the favorite in this race, though I do think he has a decent chance of winning. I want to make sure that donors are mindful of this, not because it directly affects the cost-effectiveness of their donation, but because of how donors may feel about donating $7,000, only to watch Bores lose.

And so I want to emphasize to potential donors that Bores is in fact considered the underdog. I think it’s worth donating to him despite that. But if you’re likely to feel burned by making a large donation and seeing him lose – in a way that might affect your likelihood of giving to similar campaigns in the future – then I think that’s worth taking into consideration.

What about the press coverage?

Overall, I expect that if Bores raises a lot, this will generate substantial positive press coverage. This is not to say that all of the coverage will be positive; I want to make sure that potential donors are aware of that.

I think it’s likely that there will be some unwarranted attacks made by the media on Bores’ connections to AI safety. While there are some tail risks here, my guess is that this will be basically fine for him. While fighting for the RAISE Act, Bores adeptly handled public battles with big tech, without shying away from his support for AI x-risk mitigations. I believe that he's prepared for potentially unfair media coverage, and that he will behave similarly. Overall, these considerations don’t change my bottom line.

Feeling rushed?

If you are learning about Alex Bores for the first time today, Monday, October 20th, then you might be feeling rushed into making a donation without really understanding the case and doing your due diligence.

I’m sorry about that. I understand that asking people to decide to make a $7,000 donation on the same day that they encounter the donation opportunity is a big, unfair ask.

If you feel equipped to take your time today to figure out if you want to make a donation, I encourage you to do so. And if you’d like to chat to me (or someone else) before donating, feel free to email me, and I’ll try to find time to chat with you today!

But if you don’t feel equipped to adequately assess this opportunity today, then I recommend waiting to donate. Yes, I think your donation will be about 20% less effective. However, I think it’s really important to preserve the norm of letting people think carefully and think for themselves about their donations. And I think that donations made after October 20th are still really valuable and are still the best donation opportunity that I’m aware of.

Appendix

Details of the cost-effectiveness analysis of donating to Bores

Probability that Bores loses by fewer than 1000 votes

Bores is running in New York’s 12th congressional district. This is a safe Democratic seat, which means that Bores will win the election if and only if he wins the Democratic primary. The probability that Bores loses the primary by fewer than 1000 votes basically depends on how close the primary is, so we need to understand the dynamics of the race.

Assemblymember Micah Lasher and city council member Erik Bottcher are also running for the seat. I think it’s plausible that the race will attract other prominent candidates, such as Jack Schlossberg (John F. Kennedy’s grandson). This would lower Bores’ chances of winning, but I think it doesn’t substantially affect the cost-effectiveness analysis,^[9] so I’ll be assuming that these are the only three candidates.

My guess is that Lasher is more likely than Bores to win the seat, because he was previously the policy director for New York governor Kathy Hochul and an aide to representative Jerry Nadler, who currently represents the district (but is retiring). According to the New York Times, Nadler is likely to endorse Lasher. These would be important endorsements that voters will likely care about. (I think that Bores has about a 20% chance of winning this race, to Lasher’s 50-60%.)

On the other hand, I expect Alex Bores to pick up a substantial number of endorsements of his own. Bores is considered an unusually effective legislator and has earned the respect of his colleagues in the legislature. He also has close ties to labor unions, whose endorsements matter a lot to voters.

Bottcher seems to be a relatively weaker candidate on paper (though I’d expect endorsements from some LGBT groups), but I wouldn’t count him out.

If you forced me to guess what percent of the vote (not probability of winning!) each candidate will get, I’d guess: Lasher 35%, Bores 25%, Bottcher 15%, with 25% scattered among other candidates. But of course there’s huge uncertainty. In such a race, it isn’t surprising if the winner wins by less than 1%, and it also isn’t surprising if the winner wins by 30%. A very naive model (just to get a ballpark estimate) might be something like: the margin of victory is uniform between 0% and 30%; conditional on the margin of victory being less than 1%, there’s a ⅔ chance that Alex Bores is one of those top two; and conditional on that, there’s a 50% chance that he just barely loses (as opposed to just barely wins). That would come out to a 1/90 chance that Bores loses by less than 1%.

I did some slightly more sophisticated (but still sketchy) statistical modeling that suggests a 1.45% chance that Bores loses by less than 1% (details in footnote);^[10] 1.45% is my all-things-considered guess.

Based on historical data, I’m expecting about 90,000 votes in the primary. This means that 1,000 votes is about 1.1%; so this gives a probability of about 1.6% for the event that Bores loses by less than 1,000 votes.

How much marginal funding would net Bores an extra vote?

This is the hardest part to estimate. But the heuristic I usually use is that in a primary election that doesn’t get a huge amount of media attention, campaigns can spend a marginal $100 on ads and flyers^[11] in order to attract the support of one more voter.^[12] For this race, I’m tripling that to $300, for a couple of reasons:

• I expect this to be a more well-funded race than most, because Bores’ constituents are some of the richest people in the United States (residents of the Upper East Side of Manhattan). A campaign’s 3-millionth dollar is somewhat less valuable than a campaign’s 1-millionth dollar.
• New York is an unusually expensive place to run TV ads,^[13] so dollars get you less value.

Where does the $100 number come from? Sorry, I don’t have a good source. Most online sources give numbers that are very obviously too optimistic. The $100 number comes from some combination of:

• Thinking about how much money tends to be spent on Congressional primaries and using some intuitions I’ve gathered from watching politics about how much a campaign benefits from advertising.
• Deferring to numbers thrown around casually by experts.

This means that you shouldn’t trust my numbers very much. But I have sanity-checked this number with some people I trust, and I would guess that most people reading this won’t have a better source to defer to.

This means that an extra $300,000 would better position the campaign such that Alex Bores would be able to net an extra 1000 votes in expectation, which (as per my earlier estimate) has a 1.6% chance of counterfactually winning him the election. That would translate to $190,000 for a 1% increase in his chance of winning… but earlier I claimed that $75,000 donated on launch day would translate to a 1% increase in his chance of winning. Why the discrepancy? This is because a lot of the value of campaign contributions is about signaling strength to consolidate support – as I describe below.

Early donations help consolidate support

Some of the most important things for winning a primary election are:

• Getting formidable candidates to decide to stay out of the race.
• Getting endorsements from people and groups that voters trust, such as members of Congress, state legislators, and unions.

Raising lots of money is really helpful for both of these goals:

• A potential candidate who is on the fence about running is less likely to run if they see that they’ll have to compete against a well-funded campaign.
• Politicians and unions are much more likely to endorse a candidate that they perceive as in the running, especially if they’re likely to win. That’s because endorsements are “wasted” if they’re given to candidates who don’t have a chance of winning, and also it’s because it’s useful to have good relationships with elected officials (so endorsing someone who then goes on to win is valuable).

From talking to political operatives, my sense is that money donated early on is considered much more valuable than money donated later, for this reason. It’s hard to me to put a number on how much more valuable it is, but my method was basically:

• Taking the typical numbers I’ve heard from political operatives (typically in the 2x-4x range).
• Going with the lower end of that range (because political operatives have an incentive to try to convince people that it’s urgent to donate their money now rather than later).

I think that money donated on launch day is somewhat more useful than money donated later, because campaigns typically come out with a press release announcing how much they raised on day 1. My sense is that these press releases are actually pretty important. So I’ve decided on a 2x multiplier for money donated before the end of 2025^[14] and 2.5x for money donated on day 1.

So we need to divide $210,000 by 2 and 2.5, respectively, to estimate the cost-effectiveness of donating on launch day and later in 2025. This gives us about $75k and $95k, respectively.

One last adjustment: the big tech super PAC

The more Alex Bores raises, the more likely the big tech super PAC is to get involved, and the more we should expect it to spend trying to defeat him.

I don’t have a great understanding of the dynamics here, but I would guess that there’s an:

• 80% chance that the super PAC gets involved, even if he raises very little.
• 5% extra chance that the super PAC gets involved if Bores raises an extra $1M on launch day.

In those 80% of worlds, the super PAC probably spends a few hundred thousand dollars more (let’s say $300k), though note that super PAC spending is less effective (maybe by a factor of 4) than early donations, because there is less signaling value in super PAC spending (for getting endorsements etc.), and because super PACs can’t spend in coordination with campaigns. Overall, this consideration decreases the effectiveness of giving to Bores by about 6%.

In those 5% of worlds, the super PAC probably spends $2M or so on the race. This estimate is based on how much money Fairshake, the pro-crypto super PAC that this super PAC is modeled after, spent in Congressional races in 2024. Again, we should divide by a factor of 4 or so, because this spending is less effective than hard dollars. Overall, this consideration decreases the effectiveness of giving to Bores by about 2.5%, for a total of 8.5%.

A couple of other considerations:

• Spending on Bores funges against other super PAC spending opportunities, which suggests that the 8.5% number above is a (slight) overestimate.
• But also, if Bores loses the race, then this will look good for the super PAC (some people will take it as evidence that they should be scared of super PAC spending). Though if Bores wins the race, this looks really bad for the super PAC, because Bores is considered an underdog (Micah Lasher is considered the favorite in the race, because he’s likely to be backed by the outgoing, powerful Congressman Jerry Nadler). Overall, I’d guess that this is a weak-to-moderate consideration in the direction of suggesting that the 8.5% number above is an underestimate.

Taking all of these considerations into account, I would guess that the big tech super PAC makes donations to Alex Bores about 10% less effective.

Taking our previous numbers and multiplying by 10/9 gives us about $85k and $105k, respectively – these are my final estimates.

Cost-benefit analysis of donating to Bores vs. adverse career effects

I denominate everything in milliBoreses (mB), which is an amount of good equal to a 0.1% increase in the probability that Bores gets elected.

The philanthropic benefit of donating

• If you plan to donate $7k on launch day, that has an effect of +0.7 mB, according to the above numbers.
  • But note that I expect there to be occasional (if rare) similar opportunities in the future. If your bar for donating to politics is such that your bar for donations is roughly at Bores’ level or a little lower, I expect the total impact of donating to such opportunities to be roughly 3x as good as the impact of only donating to Bores.
  • And so, if you follow the policy of donating a lot to safety-aligned political candidates, but only the very best ones, that policy will have an effect of +2 mB.
• (Since I think that non-political donation opportunities are much worse than political ones, I think it’s fair to round the counterfactual use of this money to 0-ish.)

The altruistic cost of donating

• What’s the probability that someone seriously pursuing a job in government in the future will be counterfactually denied the job because of political donations?
  • “The future” could refer to 2025-2028, or after 2028 (when we’ll have a different presidential administration). Overall, I’ll say that there’s a 50% chance that there will be a presidential administration that discriminates on the basis of political donations.
  • Conditional on that, I’ll say that there’s a 20% chance that a donation will counterfactually prevent you from being hired for technical researchers and a 40% chance for governance researchers. (There’s a chance you would have been turned away anyway, e.g. for not being sufficiently qualified or for political social media posts; and there’s a chance that your donations will be overlooked and you’ll be hired anyway.)
  • So overall that’s 10% for technical researchers. If you’re a governance researcher, we’ll multiply by 2 at the end.
• What’s the probability that you’ll want to pursue a government position? That depends on the person; we’ll call this p. (I think that for a typical safety researcher, p is around 5% or 10%??)
• How good will it be to have you in government, compared to having Alex Bores elected? We’ll call this r. (So, having you in government is 1000*r mB.) (I think that for a typical safety researcher, r is around 2%??)
• Multiplying these together, we find that the career cost of your donation is 10% * p * 1000r mB = 100pr mB.

Cost-benefit analysis

And so, it’s philanthropically good to donate if p*r < 2% if you’re a technical researcher. I think this is true for most (but not all) technical AI safety researchers. As indicated above, I think that for a typical AI safety researcher, p*r is more like 0.1% or 0.2%.

On the other hand, the threshold for governance researchers is p*r < 1%, and p is probably much higher for a typical governance researcher. So I think that for a typical AI governance researcher, donating is a pretty close call. Consider using this post to weigh the costs and benefits.

Caveats

• If you’re donating much less than $7k (e.g. if you’re only donating $500), the cost-benefit analysis may go the other way! That’s because the philanthropic benefit of your donation will be proportionally less.
• Conversely: there are good political donation opportunities that can absorb hundreds of thousands of dollars per year from each person. (Message/email me privately if you want to know more about these.) If you’re interested in becoming a really big donor, then the career costs of donating are worth it (from a philanthropic standpoint) for all but a few people, in my opinion.
• How does a possible AGI Manhattan project affect things? Well, I think that if there’s an AGI Manhattan project, then p goes up a lot for most researchers, but r is really low because there will be tons of people in the Manhattan project and steering things will be hard. So I don’t think it substantially changes the calculation. (I also think that an AGI Manhattan project is unlikely before 2029, though some people disagree.)

1. ^{^}

   It’s possible that the bill will ultimately be weakened before it’s signed, after negotiations between the governor and the legislature. I don’t have much insight into that process, though.

2. ^{^}

   Alex Bores passed more bills in his first year than any first-year legislator, and more bills in his second year than any second-year legislator.

3. ^{^}

   A budget letter is the process by which legislators officially request money be added to the state budget.

4. ^{^}

   In practice, this would mostly look like: $300-350 to persuade someone to vote for Bores instead of a different candidate (which is twice as valuable, if the other candidate is the one who beats Bores by fewer than 1000 votes).

5. ^{^}

   To be clear, I’m not making any claim like “giving the Bores campaign $8.5 million will make him win the election”. I’m just talking about the effect of a marginal dollar. And similarly, if there’s a different good thing you can do for $85,000 that doesn’t scale, we still need to be comparing the good done by 100*(that thing) to the good done by electing Bores.

6. ^{^}

   Note that I wrote a version of this paragraph before I found out that Scott Wiener might run! So I don’t think I’m biased by the recency of this news; I think that, by coincidence, we’ll be getting two amazing donation opportunities only a few days/weeks apart.

7. ^{^}

   Some green card holders have expressed concern to me that they might have problems with immigration/naturalization if they donate. The Trump administration has really shocked me with its lawbreaking; nonetheless, I think that the Trump administration is currently pretty far from being so brazen as to systematically deny people citizenship because of their political donations. I can’t completely rule it out, though; I think that donating to Democrats could have a 1-2% chance of causing someone to be unable to become a citizen. In most of those worlds, the U.S. looks much more like a dictatorship than it does now.

8. ^{^}

   This is the same number as my “early donations are 2-2.5x more valuable than late donations that can only be spent on advertising” number: 2x implies that 50% of the value is signaling and 2.5x implies that 60% is signaling.

9. ^{^}

   On the one hand, this would increase the probability that the top two candidates are separated by a small number of votes. On the other hand, it decreases the probability that Bores would be one of the top two candidates. I think these roughly cancel out.

10. ^{^}

    I used a Dirichlet distribution with parameters (3.5, 2.5, 1.5, 2.5) for the vote shares of (Lasher, Bores, Bottcher, “scattered among other candidates”), respectively. These were the values that gave the same mean vote percentages that I gave above and that seemed to me to have the correct amount of uncertainty in vote percentage. I then checked what fraction of the time Bores lost but by less than 1%; this happened about 1.45% of the time. This wasn’t super sensitive to varying the parameters of the model (within reason).

11. ^{^}

    While there are more effective uses of money for the first $100k that a campaign raises, I expect Bores to raise a lot more than that, and that his marginal dollars will go toward ads.

12. ^{^}

    In practice, this looks more like “convincing 0.6 people to vote for you instead of someone else”, because not many people will counterfactually turn out to vote just because they saw your ad. But this is about as good as getting one extra person to vote for you (instead of no one) because you’re probably taking a vote from whoever has the most votes besides you.

13. ^{^}

    A quick google search suggested that TV ads in New York are 10x as expensive. However, my guess is that in practice this means that the most effective use of money is other kinds of advertising (not TV ads). So 2x is my overall best guess about the right factor to multiply by, but this could be significantly off.

14. ^{^}

    Donating on December 31, 2025 is much better than donating on January 1, 2026, because campaigns report how much they raise in each quarter, and are judged a lot based on how much they raise in their first quarter.