Let's consider two different types of donor anonymity:
The main concern here is that PauseAI UK could be an astroturf campaign acting on behalf of, or influenced by, some interest other than the constituency that we claim to represent. For example, someone might suspect we are funded by:
An investor in one of the companies, who wants us to particularly target one of their competitors.An AI company that wants to benefit from regulatory capture.An investor who is short the AI companies, who wants us to lower their share prices by campaigning for regulation and a pause.An investor who is long the AI companies, who wants us to raise their share prices by hyping up the dangerous capabilities of AI.A foreign actor who wants to strengthen UK AI regulations to give their own country a relative advantage (even though PauseAI UK advocates for an international pause agreement).
On the other hand, there are good reasons why a legitimate donor might want to remain anonymous to the public.
They work for an AI company and they are concerned about retaliation from their employer.
[1]They also support some political candidates and they don't want those candidates to be
targeted by AI company super PACs in retaliation.They want to support think tanks or inside-game advocates who don't want to be associated with grassroots advocacy.They don't want to be approached by other non-profits seeking funding.They simply value their own privacy.
At first glance, this seems to fix all of the astroturfing concerns above. If PauseAI UK doesn't even know where our money is coming from, then we can't be manipulated by its influence.
But the problem is, how can an external person verify that we really don't know who our donors are? We could just be saying that as a smokescreen to refute allegations of astroturfing. How can the board verify that the Director doesn't know who the donor is?
And from a compliance point of view, it's generally a good idea to "know your donor". You can't take money from sanctioned entities, and if you are a registered charity there are additional disclosure requirements (although we do not currently have a registered charity, because we engage in political activity).
Finally, it might be embarrassing when you find out you were secretly backed by OpenAI.
So overall it feels like our first policy is clearly determined:
Donations above £1,000
We only really need to do Know Your Donor for amounts over £10k. But how can we know if a donor has given £10k+ to us without collecting some information on smaller donations? Donors could anonymously give a large number of smaller gifts, so we need to collect some basic info even for donations under our main threshold.
NameEmailCountry of residence
We will run a quick screen to see if the person is sanctioned, a politically-exposed person (PEP) or publicly notorious for any reason.
Donors giving above £10,000
In addition to the information above, we will ask donors:
NationalityName of the account the payment will come from.Where does this money come from?Eg. "savings from my career in software engineering", "proceeds from selling my company in 2024", "family wealth"Are you acting on behalf of, or funded by, any third party or foreign power?Will you remain solvent after making the gift?Do you have major financial interests related to AI?Broad, diversified holdings such as index funds don't need to be declared.
If someone has a major financial position in AI, should that mean we reject their donation? In most cases no – what would be concerning is a specific position that could push PauseAI towards a particular strategy (e.g. targeting Anthropic but not OpenAI). There, the board will need to ensure that such donations are unrestricted and any attempt to influence our strategy results in the gift being returned.
The danger of anonymous donations is that we might be acting on behalf of some interest other than the constituency that we claim to represent. Well, then we should have some trustworthy representatives of that constituency check our donations before we accept them, using the information collected under Policy 1.
The board of PauseAI UK is just such a group. At the time of writing the board consists of:
Joseph Miller (me), Director of PauseAI UKDavid Wood, Author and Chair of the London FuturistsJoep Meindertsma, Founder of PauseAI GlobalJonathan Bostock, AI alignment researcher at Arcadia Impact
All of our board are long-standing PauseAI members. Besides myself, none has ever been paid or otherwise compensated for their work for PauseAI. And we will make sure that the board remains such a representative and trustworthy group.
Note: These policies may change over time. We will publish any changes on our website before they are implemented.
These policies don't address everything.
Public support is our whole thing
PauseAI UK having anonymous donors normalises the idea that pausing is radical, or outside of the Overton window. We are trying to bring about a world where pausing is recognised as a common sense opinion held by everyone from concerned parents to government policymakers. It's a bit embarrassing if our own donors won't even voice their support proudly.
To give donors anonymity, you have to give non-donors anonymity
There aren't that many potential donors around. Someone might be able to figure out who our donors are through process of elimination, even if we only ever confirm who doesn't fund PauseAI. This is annoying, because we want funding advice and the first thing fundraising advisors want to know is what you've already tried. But that's the only way to keep donors anonymous.
In some sense a donor's internal decision process is their own business and they should be able to seek whatever advice they like. As long as they are the one making the call and taking responsibility for what they donate to, you could argue that they are the only person who needs to be trusted.
On the other hand, donor advisors have a lot of influence and ability to obtain funding in practice. Some advisory institutions have great sway with their donors or a large number of possible donors in their network, such that they can usually ensure whatever funding they recommend is given by someone. If one was concerned about a donation from a bad actor, it would make sense to be equally concerned about a recommendation from a bad actor.
Given that, we should probably view recommendations from major donor advisors as essentially the same as direct donations. Not from a compliance or legal perspective, but in terms of our internal processes.
Therefore, our policies applies similarly to donor advisors and direct donors. We want to know basic information about who they are and the board will decide whether to accept recommended donations.
Throughout this piece, I've tried to think from the perspective who has no trust in PauseAI UK and is looking for any reason to be suspicious. But in reality, it's abundantly clear that we do not have any ulterior motives for our work besides a desire to improve AI safety.
I volunteered for PauseAI UK for two and half years before I was ever paid to run the organisation. We have many active volunteers, the vast majority of whom will never be paid for their work.
Almost all the short-term financial incentives point strongly against us. If you're trying to make money, giving gifts to PauseAI UK is a very unlikely approach.
The best way to know that we are genuine is to just show up to any PauseAI event and talk to the people there. It will be very evident that these are real people who are there because they are concerned about AI and want to do something about it.
Ultimately, I would prefer to have fully public accounts of all donations. But beggars can't be choosers. A large fraction of the funding available to us today is only available on the condition of public anonymity.
The solutions outlined above are not perfect, but they're pretty good. At this point we urgently need to grow and hire more people. We've got work to do, so this is what we're going with.
^Do we even want donations from AI company employees? Aren't they bad people doing a bad thing?
My answer is:
It's at least plausible that some jobs at AI companies are overall useful for our cause. It's not obvious that working on AI safety at Anthropic makes the problem worse.If their motives are pure, it doesn't really matter what I think of someone, I'm not going to stop them from helping our cause.
