Home | Search

2026-02-26

Redaction policies as psychological damage

Disclaimer

• Quick Note
• Contains politically sensitive
• meta note
  • holy fuck I am sad about the fact that the general version of this problem is still unsolved after 2 years of me wrestling with it.
  • I just know a lot more specific details because I have made more contact with reality now. But the basic skeleton of this problem was sitting in my head 2 years ago itself. Holy shit.

Main

• Secrets are psychologically damaging to their keepers. They are especially damaging if either one or both of the following happens:
  • you don't have a strong moral justification for why you deserve to keep the secret
  • you don't have sufficiently many friends or a partner or just generally, other human beings you can hang out with, where other people also know the secret. You need atleast a hundred people in the know to avoid incurring psychological damage
• I don't want the fate of the entire human species and 10,000 years of history to be permanently sealed by the whoever wins the military political competition inside the US intelligence community in the year range 2026 to 2030. This is primarily true if ASI is built, although it may also be true in worlds where ASI is not built and HGE takes off or similar.
• Therefore, I am deciding to act above the law, and say that the moral justifications that people in the US IC use, don't apply to me. I need to figure out some other moral code that applies to me, if I go leak their secrets against their will.
• If I decide to leak their secrets against their will, I need to figure out a few things:
  • what moral justification applies to me. This is connected with what redaction policy I apply (and what attack targets I select).
  • what level of opsec is being followed by people in my group, i.e., who all are in the know, on the secret. This is also connected with what moral justification (and hence what redaction policy) applies to all the people inside this group. These people may need to keep this secret for the rest of their lives, depending on what the secret is.
• Simultaneously with minimising psychological damage (which is connected with moral justification and opsec), I also need to maximise effectiveness.
  • Literally no one on Earth knows what amount of secrets about the US IC are sufficient to leak, such that you could get a mass movement against them.
  • If anything, part of the reason you will not able to able to unite the public (in any reasonable span of time) is that most political groups too are busy keeping secrets about their entire lives, and engaged in culture war as a result. No one can actually relate to other people's experience beyond a point. This includes me.
  • I have a long list of guesses but no certainty:
    • Maybe we need to leak move versus countermove information in real time, to ensure the anti-ASI movement can seize power ASAP.
    • Maybe we just need to leak names and plans of all groups - both centralised and distributed - who are attempting to build ASI. This group will expand with time as GPU proliferation continues.
    • Maybe we need to leak names and plans who support human extinction
    • Maybe we need to leak names and plans of people who are making moves and countermoves against democratic safeguards (which is a separate but related thing from people who are making moves and countermoves against the anti-ASI political movement)
    • Maybe we need to leak names and plans of people who are personally abusive, like those who are rapists or similar
• Specific ways I can pull off the plan of leaking US IC secrets against their will:
  • Supporting whistleblowers
    • Find someone who has risen in ranks of UC IC to the point where they know important secrets, and are part of the few hundred people in the know. But, they no longer buy into the moral justification of keeping those secrets.
    • Redaction policy - In my whistleblower guide, I have given the whistleblower two options:
      • Decide your own moral code and redact information yourself. - The whistleblower is undergoing significant psychological damage because their opsec is bad (they don't have a large group supporting them) and their recent moral justification for their entire life just broke. It is a tall order to expect them to spend months carefully redacting info. It is an even taller order to expect such a whistleblower to invent a new moral code and corresponding redaction policy from scratch.
      • Hand over everything to some journalist and trust their moral code. - The billionaires funding the mainstream media houses in the US are increasingly captured by the UC IC themselves, and the other journalists are increasingly far-left people engaging in class war. "My class deserves to keep secrets, your class does not."
    • Both of these moral codes seem bad, and I personally want nothing to do with them.
    • My whistleblower guide does not mention this topic at all, because I don't yet have a confident answer to this question.
  • Supporting cyberhackers
    • I need to hire a group of hackers, while simultaneously maintaining independence from the russian or indian IC, to avoid getting sucked into their own nationalist moral justifications.
      • The best way to avoid getting sucked in is demonstrating competence and alignment of interests. The best way to get this level of competent hackers on our team is to have a better moral code that appeals to more hackers.
    • Opsec - If I don't want people in our group to undertake psychological damage, then ideally our group must have a few hundred people. Or atleast, the subset of our org that is in the know on the secrets must be atleast a few hundred people. It seems hard to run such a large group and yet maintain this level of independence in terms of moral justification believed by its members.
    • Otherwise, I just keep thinking of new ways to pass the buck to someone else:
      • We could just not look at the info we acquire, and blindly pass it on to some journalists. See above for why most journalists' rationalisations are shit. Either they are captured by UC IC or non-US IC themselves, or they are busy engaging in class war.
      • We could have a few people in the group who are designated redaction specialists, and ensure only they take on all the psychological damage. If I am running such a hacker group, I have to belong to this subgroup too, and hence undergo damage myself.
      • We could leak everything from the selected targets whose info we acquire with no redactions. This sounds good in theory but makes me feel bad the moment I put it into practice. I actually have no way to predict in advance whose all secrets are sitting on the hard disks of UC IC servers, it is likely not just info about their agents but also all sorts of unrelated info about the public.
      • We could make attack capabilities so strong that we and our allies can hit all the targets on earth all at once, and leak all the world's secrets with no redactions. This is the harvey dent fair coin option. There is no way we can possibly get informed consent to do this, because that requires solving the culture war itself, which is hard because various groups in the culture war are also busy keeping their own secrets as part of the propaganda war they are waging.
        • I am still looking for some "middle plan" between we leak very little about some specific actors (like Snowden did) and this maximal plan. I don't yet actually have such a middle plan ready.
      • We could have an AI do the redactions so that we need less humans (maybe even zero humans) to see the unredacted secrets. This sounds like an elegant solution but it still requires us to all agree on some redaction policy to prompt the AI with.
      • We could just kill the people doing the redactions after they're done with their job. This way we don't have to even think about opsec and mental health of these people. Getting people to voluntarily sacrifice their lives in this way is also very hard, and requires a strong moral narrative.
    • Where do I hire these hackers from?
      • It is not sufficient for me personally to have a moral code and redaction policy I believe in, I also need everyone else in the hacker group to also abide by this code. This is hard (but might still manageable), because most hackers come from some existing life experience and moral code already.
      • US cyberhackers are already either contracted by the US IC (via some chain of subcontractors), or they're busy doing regulatory theatre for the S&P500 companies, or they're very libertarian-leaning, or they're far-left class war people, or they'll destroy literally anyone's life for sufficient money for themselves or their family, and so on. Russian cyberhackers are the same, except that they're even more skeptical of moral narratives as compared to US, and more interested in money. They have more rationalisations around why not just govt but entire society is corrupt, and why moral narratives are a cognitive weapon wielded by bad people. (I need to get more data to confirm this hypothesis)
      • If I invent a new moral code and redaction policy, I also need to convert a bunch of hackers into this moral code. This too is hard and takes time - time that I may or may not have.
• What is my personal moral code if I ignore all the leaking stuff?
  • I have noticed that thought experiments do a bad job at this. It is easy to press a button or pull a lever. It is easy to give orders for someone to kill people on your behalf. See also: common failure mode of hiring hit men irl. It is much harder to spend 5 years constructing a button that enables mass murder. Power buys you distance from the crime, as Elizabeth puts it. Morality is decided in action not thought.
  • I want to kill atleast one person. I want to kill the minimum number of people required to stop ASI from levelling the planet in the next 5-10 years. See the other section on longterm invariants (turing-complete invariant, political system as invariant, and short-term stopping ASI to build an invariant).
  • I probably can't compute this exact minimum number of people in advance because I am not (yet) a god. I think I would especially feel bad if I had killed multiple orders of magnitude more people than was absolutely necessary to stop ASI in the next 5-10 years.

Side note

• making contact with reality
  • I have to go talk to more hackers on this
  • I have to get some actual datasets and experiment with applying different redaction policies on it
• the fact that I am still stuck after 2 years is all the more reason I need to go make more contact with reality on this topic (though it is hard, most people want nothing to do with a discussion around cyberattacking the US govt. ideal discussion partners are either people in india/russia who have no future intention of going to US/UK/EU, or the crazy radicals like myself who are sold that they should turn their life upside down just to work on this, and will even fly to india just to chat with me on this topic)

Subscribe

Enter email or phone number to subscribe. You will receive atmost one update per month

Comment

Enter comment