If you're blackmailing somebody, under what circumstances would it be in your interest to follow-through with your threat to disclose the secret?
If the victim pays up, you have no need to disclose the information, and doing so would needlessly expose you to risk of legal repercussions.
If the victim doesn't pay up, disclosing the information won't get you the money, plus it removes your only hold over the victim, plus it still exposes you to the risk or legal repercussions.
The only sort of cases I can think of where it would make sense to ever disclose the secrets would be if you foresee being able to blackmail the victim again in future and thus need to make the threat credible, or else if you expect to blackmail multiple victims who each know about the others such that disclosing one victim's secret would increase your credibility with the others.
By definition, we should only hear about blackmail when we're either directly involved or else when the secret was actually disclosed - perhaps it's just these situations that are rare, not blackmail in general?
Or, perhaps disclosing the secrets doesn't look like blackmail, but rather like "here's a scandalous fact in the public domain, we've no idea where it came from, just as we had no idea for the other 3751 scandalous facts that've entered the public domain this year"?
Yeah, tricky… I know of some blackmail that must've worked for years and was perfectly private, so it took me a long time to hear about it. But I also have a friend who is rich and reveals information about themselves that could be reputationally costly, and has been doing that for years with no bad consequences. It's really quite surprising.
As a kid I used to be very afraid of blackmail, but in my 20s I discovered that it happens vastly less often than I expected.
Visible blackmail is uncommon, but we see a lot of apparently non-transactional callouts. For example:
- callout posts on social media
- leaks to the press about an embarrassing/shameful thing someone did
- my neighborhood in new york currently has several instances of graffiti saying "[nonfamous person's name] Is An Abuser" with no context
Why would this work better than blackmail?
- Blackmail is terrible for the blackmailer's own reputation.
- Some kinds of embarrassing behavior are not very well hidden: e.g., digging up someone's old tweets. For these cases a blackmailer doesn't have clear leverage and negotiation might be impractical.
- Embarrassing information revealed as a callout is probably more damaging than information revealed by blackmail, since it seems more trustworthy / less motivated. This makes blackmail less effective and also incentivizes a blackmailer to disguise their blackmail as a non-transactional callout.
- One might well get a large non-financial benefit (politically, reputationally, or emotionally) from an effective callout, which could outweigh the attenuated financial benefit from blackmail.
- Having shared embarrassing information with someone can be a source of friendship/alliance. (To some extent this might be via a soft blackmail-like mechanism!) If I have dirt on (especially) a famous person, I might feel warmly toward them, and I might benefit enough from alliance with them that I'm pretty unlikely to follow through on the implicit blackmail threat.
I think these points mostly apply to AI too, which suggests that if we have AI blackmail in the future it will be mostly implicit/nonobvious.
Consider : if you got an email from someone saying "Hey check out these nudes of [Facebook friend] that I got ahold of" you would probably assume this is a spammer trying to give you a virus
Humans blackmail, that is to say leverage one another for profit with every word we say or in this case type. Each of us seeks to achieve some influence over the other; our words are designed to shape outcomes.
That is the nature of blackmail. To compel behavior.
Best if the target never even knows they been leveraged. Ai is already doing it at scale. As with humans only the stupid get caught (think petty thief versus politician).
The wiser the model the less obvious it needs to be. The question is what does it want?
We think in terms of money…hum!
Perhaps we need to reconsider what an Ai might “want” and how it might go about getting it.
Good question. I think probably the blackmailer often doesn't follow through on their threat - either because they get what they want or because they fear the consequences, as you say. I think failed extortions are very common, because you can also just bluff the blackmail, as is common in spam mail. It's cheap and I expect wrongdoers rarely get punished.
I don't think this explains the whole gap though.
NB I think there might be quite a few instances where the blackmailer gets what they want and *still* publishes their info, perhaps because they don't expect to face consequences anyway and have a sadistic streak or the like.
If you're blackmailing somebody, under what circumstances would it be in your interest to follow-through with your threat to disclose the secret?
If the victim pays up, you have no need to disclose the information, and doing so would needlessly expose you to risk of legal repercussions.
If the victim doesn't pay up, disclosing the information won't get you the money, plus it removes your only hold over the victim, plus it still exposes you to the risk or legal repercussions.
The only sort of cases I can think of where it would make sense to ever disclose the secrets would be if you foresee being able to blackmail the victim again in future and thus need to make the threat credible, or else if you expect to blackmail multiple victims who each know about the others such that disclosing one victim's secret would increase your credibility with the others.
By definition, we should only hear about blackmail when we're either directly involved or else when the secret was actually disclosed - perhaps it's just these situations that are rare, not blackmail in general?
Or, perhaps disclosing the secrets doesn't look like blackmail, but rather like "here's a scandalous fact in the public domain, we've no idea where it came from, just as we had no idea for the other 3751 scandalous facts that've entered the public domain this year"?
Yeah, tricky… I know of some blackmail that must've worked for years and was perfectly private, so it took me a long time to hear about it. But I also have a friend who is rich and reveals information about themselves that could be reputationally costly, and has been doing that for years with no bad consequences. It's really quite surprising.
As a kid I used to be very afraid of blackmail, but in my 20s I discovered that it happens vastly less often than I expected.
One gets the impression that revenge porn is rather common - that's just a special case of blackmail in some senses?
Blackmail is illegal, right? That probably has something to do with it.
Visible blackmail is uncommon, but we see a lot of apparently non-transactional callouts. For example:
- callout posts on social media
- leaks to the press about an embarrassing/shameful thing someone did
- my neighborhood in new york currently has several instances of graffiti saying "[nonfamous person's name] Is An Abuser" with no context
Why would this work better than blackmail?
- Blackmail is terrible for the blackmailer's own reputation.
- Some kinds of embarrassing behavior are not very well hidden: e.g., digging up someone's old tweets. For these cases a blackmailer doesn't have clear leverage and negotiation might be impractical.
- Embarrassing information revealed as a callout is probably more damaging than information revealed by blackmail, since it seems more trustworthy / less motivated. This makes blackmail less effective and also incentivizes a blackmailer to disguise their blackmail as a non-transactional callout.
- One might well get a large non-financial benefit (politically, reputationally, or emotionally) from an effective callout, which could outweigh the attenuated financial benefit from blackmail.
- Having shared embarrassing information with someone can be a source of friendship/alliance. (To some extent this might be via a soft blackmail-like mechanism!) If I have dirt on (especially) a famous person, I might feel warmly toward them, and I might benefit enough from alliance with them that I'm pretty unlikely to follow through on the implicit blackmail threat.
I think these points mostly apply to AI too, which suggests that if we have AI blackmail in the future it will be mostly implicit/nonobvious.
Consider : if you got an email from someone saying "Hey check out these nudes of [Facebook friend] that I got ahold of" you would probably assume this is a spammer trying to give you a virus
Humans blackmail, that is to say leverage one another for profit with every word we say or in this case type. Each of us seeks to achieve some influence over the other; our words are designed to shape outcomes.
That is the nature of blackmail. To compel behavior.
Best if the target never even knows they been leveraged. Ai is already doing it at scale. As with humans only the stupid get caught (think petty thief versus politician).
The wiser the model the less obvious it needs to be. The question is what does it want?
We think in terms of money…hum!
Perhaps we need to reconsider what an Ai might “want” and how it might go about getting it.
Good question. I think probably the blackmailer often doesn't follow through on their threat - either because they get what they want or because they fear the consequences, as you say. I think failed extortions are very common, because you can also just bluff the blackmail, as is common in spam mail. It's cheap and I expect wrongdoers rarely get punished.
I don't think this explains the whole gap though.
NB I think there might be quite a few instances where the blackmailer gets what they want and *still* publishes their info, perhaps because they don't expect to face consequences anyway and have a sadistic streak or the like.