I’m close to a single-issue voter/donor. I tend to like politicians who show strong support for AI safety, because I think it’s an incredibly important and neglected problem. So when I make political donations, it’s not as salient to me which party the candidate is part of, if they've gone out of their way to support AI safety and have some integrity.^[1] I think many people who focus on AI safety feel similarly.

But working in government also seems important. I want the government to have the tools and technical understanding it needs to monitor AI and ensure it doesn’t cause a catastrophe. Some people are concerned that donating to Democrats makes it harder to work in a Republican administration, or that donating to Republicans makes it harder to work in a Democrat administration. Administrations understandably care about loyalty (though they also care about domain expertise), and they have to filter through a lot of people and must make quick heuristics-based judgements. So even if you vibe with the political philosophy of the administration, if you donated to the other side (even if only because a few of their politicians supported AI safety), your donation can make it hard to get certain government jobs.

(I think strong versions of this concern might be a little like giving into blackmail, and I’m very sympathetic to arguments that people should ignore this concern and donate to whatever politicians they want to. I’m considering that question out of scope and aiming this document at people who would ordinarily be too hesitant to give to politics at all. I want to improve on that baseline.)

The reasoning goes something like “if everyone knowledgeable about AI safety donated to every politician who supports AI safety, then fewer people would be able to work in government.”

But if everyone knowledgeable about AI safety refused to donate to politicians, then the politicians championing this important cause would be left high and dry (and fighting against some large super PACs punishing them for supporting safety).

In this post I'll summarize my understanding of how real the career concerns are after talking to several experts and people with first-hand experience, and I'll propose a course of action that I think lets people get as much of the best of both worlds as possible.

This document is largely aimed at technical people who could do valuable specialized work in the government, since many technical people don’t have strong partisan views and are thus open to working for any administration but don’t know how to navigate the relevant career considerations.

Background/facts

• Direct contributions^[2] to candidates’ campaigns and PACs are public record. For example, you can search people's names on this site and see the donations they have made.
  • It’s particularly easy to find federal donations but even state-level donations are public.
• Many people who have gone into policy work regret having made political donations in the past.
• Most government roles are civil service roles. These are the ~2.4 million people who do the day-to-day work of the federal government. They are not very political positions, and people typically build a career in these roles, continuing to serve through Republican and Democrat administrations. Traditionally, the government doesn't look at political donations when hiring for these roles.
• The Trump administration is placing an unprecedented amount of emphasis on loyalty and political affiliation in civil service roles. There is speculation that the Trump administration in particular might be testing for party loyalty via political donations when hiring for civil service roles. (This is abnormal and illegal.)
  • This is relatively unprecedented and a new feature of the Trump administration, but it is possible that future administrations might copy what the Trump administration is doing.
  • The Trump administration in particular also cares about loyalty to Trump more than just loyalty to the party. So for example they also dislike donations to Republicans who have been adversarial to Trump.
  • Some people seem under the mistaken impression this filtering is a large factor or is obligatory. It is not. I don't think it's being checked systematically. I think it's a larger issue the more prominent/leadership-y a role is. Overall I estimate donating >$10k to Democrats makes it ~20% harder to become a civil servant right now. I think not having a security clearance is about as costly as having donated to Democrats for civil service roles (so consider trying to get one of those if you want to keep those doors open).
  • The Trump administration has also created a new class of civil service role called a Schedule Policy/Career appointment.  There are ~50,000 of these roles and they're more like civil service roles without the traditional protections against partisan screening.^[3]
• The other kind of government job is a political appointee role. These are a handful of political positions where you work with the President and his staff to administer his agenda. They only last as long as the administration and traditionally screen heavily for political alignment with the administration.
  • These roles are things like leadership level of federal agencies, Cabinet secretaries, agency heads, and senior advisors, etc.
    • There are only ~4000 of these roles (and only a tiny fraction of these have any relevance to AI safety): around 1,300 Senate-confirmed appointees (typically the highest ranking executives of government agencies); around 500 appointees not requiring Senate confirmation (most commonly White House and related roles); around 750 members of the Senior Executive Service (senior management positions in federal agencies), and around 1500 Schedule C appointees (more junior policymaking roles, not leadership).
  • When hiring for these roles, the Presidential Personnel Office (PPO) will also look at your political party affiliation, public statements, and try to test for party loyalty.
  • They don’t seem to care about those other factors as much as donation history, though. If you’ve made statements the PPO might dislike but those statements aren't easy to dig up, they might not find them.
  • Even tiny donations matter.
    • I've heard one story where a $25 donation to Hillary years ago prevented someone from getting a fairly important political appointee role. I’ve heard many other stories that rhyme with this. So far I haven't heard any stories like this for positions in civil service (as opposed to political appointee roles).
    • However, it seems like small donations are surmountable, especially if they weren't to a direct opponent of Trump. Bigger donations are worse and are harder to surmount, but you should not write yourself off for ever working in civil service in the Trump administration if you've made a large donation to Democrats as an adult, particularly if you might be an amazing candidate for other reasons.
• Sometimes, but not always, roles as staffers for congresspeople and occasionally think-tank roles will consider at your party affiliation and donation history when deciding whether to hire you.
• For technical/engineering civil service roles (not leadership), it seems like political donation history doesn’t matter that much. There’s been a little fighting about it but it mostly hasn't affected things.
  • I believe this based on checking with someone I trust working in the federal government who has dealt with similar situations in the Trump administration.
• In terms of getting government roles, there is limited upside to donating. For civil service roles they don’t preferentially hire people who have donated, and for political appointee positions donations are just one of many ways to establish connections to the party (except maybe if you’re giving more than $1M or so).
• Donating to both parties does not make you neutral, and it is hard, though not impossible, to make up for donations in the past to the wrong party.
  • To make up for donations, you would probably need:
    • A good story where you had a change of heart
    • To have never changed parties before
    • To donate ~5x more to the new party
• There are some important political appointee roles right now where people in the policy world claim it has been hard to find someone qualified to fill them because all of the best candidates have donated to Democrats. Mostly these seem to involve people who had a lot of relevant credentials and experience.
  • I think in policy land, it's surprisingly rare to find people who have not made donations because everyone in policy land is politically active and cares about politics. In my experience in technical land, it is much less rare to find people who have never made political donations because many technical people don't really follow politics. This is unfortunate because the technical people should be donating, and the policy people should not. But I think it gives policy people an impression that people who have never made a donation are very rare. When, if you are a technical person, this is not as much the case.
• If Vance is elected in 2028, it is hard to predict exactly what his policies might be. They might differ from Trump's in some ways, so there is some general uncertainty.
  • Trump also might change his policies at any point, in any direction

• I know of no examples where an individual’s donation history^[4] affected their ability to run an organization that wasn’t directly working with the government, or affected the public’s perception of that organization.

  • Outside of partisan organizations like political advocacy groups or think tanks, it is very rare (and in many states illegal) for employers to decide whether or not to hire people based on political affiliation/donations.

  • There has been a small amount of backlash towards EA in general due to political involvement, but this was for much more out-of-distribution involvement than political donations.

• Specific "covered associate" jobs in banking/finance require employees to not make or have recently made large or non local political donations.
• Some think tanks need to maintain neutrality, so they won’t accept donations from people who seem partisan (esp. if they seem left-wing).
  • This is mostly a concern if you’re considering donating at least a million dollars within your lifetime.
• Right now, public affiliation with EA and AI Safety might be a bigger barrier to working in government than your party affiliation.
  • This is true in the current administration, was true in the previous (Biden) administration, and may or may not end up being true in the future or in different administrations, it’s hard to predict.

Recommendation

Obviously, this recommendation is kind of silly, for example, in reality some factors multiply instead of adding, but I think that when people see a document with a whole bunch of considerations and no way to aggregate them together, they just get overwhelmed and bounce off.

I think putting them in this format forces people to actually make decisions and wrestle with the magnitude of various numbers. I expect this algorithm to give a better answer than if you just sat down and tried to think about things for 30 minutes, unless you have pretty unusual circumstances that are not covered by the algorithm.

1. If you plan to donate less than $500 to politics in a year, don’t bother. (This has been touched on in another post.)
2. If you're planning on working as a covered associate in a financial institution in the next 2 years, or you plan to take on the role of a nonpartisan public intellectual,^[5] don't donate to politics.
3. If you can only picture yourself working with people from one particular party, donate to candidates from that party only.
4. If you are actively seriously considering a career in policy and open to working in a Republican administration, or are still in college and unsure what you want to do with your career, do not donate to politics at all unless:
   1. You would otherwise give at least $10k to politics over 2 years (if not for these career considerations) and you are not interested in political appointee positions (which are pretty rare, e.g. the highest ranking roles in federal agencies) in a democratic administration. If that’s you, donate only to Republicans.
5. If you are planning on donating more than $1M over your lifetime to specific think tanks that might not want to accept money from people who have funded Democrats, you may not want to donate to Democrats. If you are interested in learning more about donation opportunities like that, DM me.
6. Even if the above bullets don’t apply to you, if you are an exceptionally good fit for working in a Republican administration (95th percentile or higher among people who are not actively seriously considering a career in policy): (1) do not donate to Democrats. (Donating to Republicans is fine.) (2) consider strongly switching to a career in the Trump administration right now (e.g. fill out this advising form).
   
   To test if this is you, see if you get at least 11 points in the following quiz:
   • Do you have security clearance? +4

     • If not, do you expect to be able to get security clearance (US citizen, not dating a Chinese/Russian/Iranian national, doesn’t use illegal drugs or could easily stop using illegal drugs^[6])? +2

   • Are you willing to work in the Trump administration? ($100k+ pay cut, less job stability or freedom, likely moving to DC, etc) +3 (If yes, also consider just jumping into politics directly! Fill out this form!)
   • Can you get emotionally hyped up and excited about Trump’s policy agenda?  +2
   • Can you work professionally and respectfully with people across the political spectrum and not come across as horribly weird and/or abrasive? +1
   • Can you fit in in a Republican culture? +1
   • Have you… (subtract all that apply)
     • Ever donated to Democrats since turning 18 OR donated to a prominent anti-Trump Republican? -1
     • Donated to Hillary Clinton, Joe Biden, or Kamala Harris? -2
     • Donated at least $1000 to Democrats? -1
     • Donated to Democrats in the last 4 years? -1
     • Donated at least 5x as much to Republicans as Democrats since last donating to Democrats (or are you planning/willing to do that in the near future)? (This only applies if you’ve donated to Democrats) +1
     • Been registered as a Democrat? -0.5
     • Made public statements that someone could dig up that sound woke or anti-Trump? -0.5
     • Made public statements that are very easy to find (such as in a newspaper) and are very woke or very anti-Trump? -1
   • Are you obviously/publicly EA or AI safety affiliated? -1
   • Have you worked at a strongly EA/AIS-affiliated institution? -1
     • Is that organization particularly “doomer-coded”/not widely respected as doing technical work? (You can ask your employer about this.) -1
   • Are you a Republican at heart? +2 (If yes, also consider just jumping into politics directly! Fill out this form!)
   • Do you have a good understanding of and respect for Republican culture, writing, and ideals? +2
   • Do you have connections in Republican circles (for example, the tech right), or do you actively plan to build those connections soon? +1
   • Do you have the credentials of someone who has worked in industry? Eg have you worked for at least 2 years in a big AI lab? +1
   • Are you generally very competent and reliable all around:
     • Are you good at responding to emails, making meetings, being in the office on time every day when it matters, and not silently dropping professional commitments? +1
     • Do you produce high-quality output reliably? (Obviously this varies by job. One bellwether is if you’re a researcher, do you have enough output that you could give a talk on your work every quarter to your peers where they would feel like they learned important interesting things?) +1
   • Do you have a track record of orienting in confusing epistemic situations and agentically finding creative ways to make things happen, even in the face of tight constraints? +1
   • Seniority (check all that apply)
     • Are you at least 30 years old? +1
     • Are you at least somewhat senior? For example, could you be a MATS mentor? Do you have a PhD? Or similar. +1
     • Are you very senior? For example, do you lead an organization? +3 (If yes... you guessed it! Fill out this form!)
     • Can you imagine there’s at least a 3% chance you’d lead an important government research center or standards project at some point in the future? +2
   • Do you have relevant credentials in policy circles? Have you worked in policy before, authored something like a well-regarded think tank report or legislation, etc? +1
   • Are you a US citizen? +1
   • Is your spouse or potential future spouse…
     • already in a position in the admin? +2
     • OR alternatively seriously considering a position in a Republican administration? +1

This check list is not comprehensive. If you feel like there’s something unique about you that this list doesn’t capture, use common sense to adjust your score. I also recommend talking to an advisor if your score is 8 or above and you could feel really excited about working in the admin! (I know I’m plugging that a lot, but all of this is just very individual so the value of talking to an expert is high.)

I am aware it’s hard to score an 11 on this quiz and that it filters most people out! That’s the point!

Explanation

• I think some people want to preserve option value in case they “get tapped”/a super impactful DC opportunity falls from the sky. I think this is the wrong mental model of how things work. I think the DC opportunities that are most impactful only come up after you’ve established yourself in DC and built up some connections and career capital. If you’re seriously open to/interested in a potential career pivot, you should try building that foundation now. You can do that part-time while keeping your current job. One first step is to fill out this career advising form.
  • People I know who got swept up into good government roles ~all tried establishing themselves in policy circles before they got their roles. For instance, people point to Dean Ball as an example but he was a prolific policy writer and had a great network. He had a very traditional conservative career (Manhattan Institute, Hoover Institution, Mercatus) -- one which did not keep his option open for a Democrat administration.
  • Establishing yourself in this way only takes 6 months-12 months, so it’s very achievable, but you need to put energy into pivoting before you get the role/before the career switch is de-risked.
  • If you want to go into government, making that pivot now is probably better than in the future, since I generally expect career capital and trust to accrue over time and I think many organizations/initiatives could get established now that could grow over time (and I think it’s very hard to grow things quickly even if there’s political will for it).
  • I think there are some exceptions to this if you are very, very senior or if you are deeply aligned in your heart and culture with the politics of the current administration.
• Overall, the future is uncertain and hard to predict. So I think that some technical people might end up needing to go into government. Thus some people should probably preserve their option value. But instead of having people randomly decide whether or not to donate, it seems better to have only  the people who would be the best fits for working in government hold off on donating and be ready to jump into those roles if needed.
  • I think people can identify whether they’re a good candidate for working in government in advance, and I think that it's unlikely that it would make sense for most technical researchers to go into government at any point. (Even in worlds where there’s a need for a lot of people in government, in most of these worlds, there is still a lot of science/engineering to do, and someone has to do it.)
• It seems valuable to have some very senior people who can fill the very occasional roles that come up that are important and require a very senior person with a deep background in AI and its risks.
  • I think there might be a need for 1-5 people to step into a role like this at some point in the future. (That’s an 80% CI.)
  • Thus, having 70th percentile+ very senior people who are good fits for going into government preserve option value seems good.
  • I say 70th percentile because there might be other constraints that are hard to predict on who would be good fits for these roles. And because there's a small enough pool of people that if a few people drop out, I want to have backups.
• I think it's good to have a small number of people who are very well-suited to working in government ready to jump in, in case they need to. For example if the government wanted to hire many more people to grow existing branches that keep tabs on AI, or create new agencies for that.
  • I think it is good for 95th percentile+ fits for these kinds of roles to not donate so they can take these roles. I think this is fairly easy to predict in advance.
  • It's hard for me to imagine a world where it’s crucial for more than 40 people (i.e. ~5% of qualified people) to fill a role like this and where the government is being very picky about who could be hired.
• The political donations of your spouse can also affect if you’re hired for political appointee positions, though this effect is weaker and only important for heavily scrutinized positions.

Is it possible you will end up working in government?

Many policy positions can be very impactful. If your career plan is to do policy work, then you already know the answer. This is for people who do not currently plan to work in policy/government.

• Maybe you want to work at CAISI at some point in a leadership-y role (or another org like CAISI that comes to exist in the future). Note that many government organizations aren't able to hire right now but that could change at any time.
• Occasionally (~once a year?) roles open up that would be good for a very senior/credentialed technical person who has a deep understanding of AI and its risk to fill. These roles are sometimes fairly influential.
  • People who run organizations and have pretty legible careers seem like the right fits here.
• Maybe in the future, the government will nationalize the labs, or there will be some kind of important national project to build AI, or there will be an emergency and lots of AI and AI safety people will be fast-tracked into the government.
  • Again, almost all these roles would be civil service roles, not political appointee roles.
  • If this happens, it seems like there might be more tolerance to overlook things that would be disqualifying right now, since it’s hard to imagine quickly hiring many talented capable people.
    • This was the case for the Manhattan Project, for example (Oppenheimer had known, strong ties to the Communist Party).
  • In this world, there is probably still a bunch of valuable stuff to do that is not inside the government.

Other advice

• If one party or the other speaks to you more, consider specializing in donations to that party and not donating to the other side. This keeps your options open for political appointee positions in your party of choice. If everyone followed this strategy, both parties would get donations and people would still be able to go into political appointee positions for their party of choice.
  • Though there is some cost to doing this because it means you would end up giving half as much to the very best opportunities for each party, as opposed to the next-best things.
• If you’re a fan of a particular political party and already making donations to people in that party, consider giving to more mainstream candidates in that party, not just people who are sympathetic to AI safety. I think it can come across as very one-track-minded and weird to only support a small number of people very strongly.
• You might want to ask your peers for their opinions on political donations.
• Specializing in donating to Republicans seems like a much safer option than specializing in donating to Democrats.

Counterarguments

These are some counterarguments I’ve heard from knowledgeable experts who have disagreed with parts of this document:

• Maybe you think political donations aren't very effective/good. For instance you might think the executive branch is much more powerful than the legislative branch, so donations matter much much less than working in the administration. This writeup isn't trying to analyze the benefits of political donations, only the costs, so adjust your thresholds appropriately if you think the benefits are extremely large or small.
• Maybe you think that technical work is not promising/tractable, in which case working in government looks relatively much better and maybe many more people should go into government.
• Maybe you think the unknown unknowns are too great and the stakes are too high (though not making donations also has consequences!).
  • For instance maybe you think there will be a time when the government wants to hire many people to do very important work but they don’t in fact relax their hiring requirements.

Rambling

• I think it's easy to feel like your donations don't matter. And then when the career effects do crop up, those feel like they matter because they're very salient. But donations really do matter, and the upside is real. I'm worried people will be too risk-averse.
• People seem to come up with quite a lot of reasons not to donate to things in general, many of which I am fairly skeptical of. I do think it is a tough decision. But, I see a lot more creative reasons not to donate than to donate.
• Lots and lots of people seem to have some story for why they really need to remain neutral, even if they have no plans of going into government. I'm pretty skeptical of these stories, and I notice people don't take other actions that would be in accordance with these plans to remain neutral when those other actions aren't as costly and annoying as donating money (for example I never see them advocate for others in similar positions to remain neutral).

How did you come up with these heuristics?

I am a technical person, not a policy person. So take everything I say with a grain of salt. I have talked to ~7 different policy people about this stuff and ran this doc by some of them.

More than once I've seen different reviewers say my bar is too high or too low for the exact same item in my list, so I think I’ve successfully synthesized everyone’s positions fairly well. But keep in mind people disagree a lot!

1. ^{^}

   Though I also care about other things; it’s a bit of a simplification that I’d necessarily automatically donate to someone just because they sponsor one bill.

2. ^{^}

   People sometimes also support candidates through means other than directly donating to the campaign. For example super PACs (not to be confused with PACs), or indirectly via c4s (which can’t advocate for a candidate but can do things like targeted voter registration efforts). I don’t recommend doing this as these opportunities are less impactful (>5x worse, except in rare cases where the campaign is being very silly or an external org is being very clever), only apply to certain races, and are often already filled by larger donors.

3. ^{^}

   It's worth noting this could have some real benefits. The argument for them is these protections previously made it burdensome to fire federal employees who weren't performing well. I haven't personally looked into if they're a good or bad idea overall.

4. ^{^}

   Outside of very rich people creating super PACs and sinking many millions into elections.

5. ^{^}

   Which is not to say I think nonpartisan public intellectuals are particularly better than partisan ones; both seem useful.

6. ^{^}

   I suspect people sometimes overestimate how hard it would be to get a security clearance if they currently use illegal drugs. While I’d very much recommend not using any illegal drugs if you plan to get a security clearance, I think in practice you have a good shot at getting your clearance anyway even if you haven’t been perfectly clean for a year, as long as you’re honest about it, your usage has been quite moderate, and you are capable of staying clean going forward.