That's not as narrow as possible, that's narrow enough to satisfy strict scrutiny. Those are not the same thing. You could make a narrower law than what is necessary to satisfy strict scrutiny. (For instance, as Rocky mentioned, by limiting liability to only the user).
The reason I bring this up is, as I mentioned, there are laws that satisfy strict scrutiny, which people on Techdirt don't support. That is a necessary but not sufficient criteria.
A relevant example that I already mentioned is 230. 230 would still preempt a version of this law that satisfied strict scrutiny via the proposed fixes, and Kathryn'sthreadexplicitly mentions this.
The whole point of this law is it introduces liability for the platform for content a user produces. If that's your stance, you're saying you still wouldn't support a version of this law that passes strict scrutiny.
That is a defensible stance, but is explicitly not the “if this law just fixed strict scrutiny issues, I’d support it" stance that is being implied. It's the other one.
Combined with Mike’s suggestions for how to write a better law targeting that bullshit, the article implies that he would be in favor of such laws so long as they’re narrowly tailored for that kind of content.
How do you square that with his stance that liability should always be on users? Or how strict liability can lead to companies not moderating, to avoid actual knowledge? Genuinely, I can see that going either way. There's "this law has strict scrutiny issues and other problems", and there's "if this law fixed strict scrutiny issues, I'd support it". Either way is defensible, but I don't think you can guess it from the implication.
And for the record: I hate this nudification bullshit too, but I’m a strong believer in crafting speech-adjacent laws as narrowly as possible. Minnesota could have done a better job here.
I agree on that, I just want to know what 'narrowly as possible' entails, since we tend to disagree on that. Usually your guys' bar on this sort of thing is that it has to qualify as first party speech. Maybe this is an exception, but I wouldn't assume either way
Addendum: I also forgot to mention that a law that satisfies strict scrutiny doesn't necessarily satidfy your long time stance of users themselves are the ones who should be liable for any violative actions they make rather than the platform, either
I explained why and how Minnesota could have written a law that would pass strict scrutiny,
You did, but you also did not actually say whether you'd support such a law. Passing strict scrutiny and being a law you'd support are not the same thing.
For instance, as I mentioned, there are other barriers besides strict scrutiny, which you're aware of. 230 preemption is a legitimate problem for any sort of state level civil enforcement.
which you then falsely claim is me stating that I wouldn’t support a well written law?
Nowhere in the article do you actually state one way or the other whether you would support such a law. I'm not claiming you would or wouldn't, I'm saying the wording you picked doesn't actually say which. You're smart enough and write carefully enough that that rings alarm bells for me.
It matters, given that your bar for supporting speech laws in general is not solely whether they'd pass strict scrutiny or not. There's plenty of laws that pass strict scrutiny you (reasonably) still do not like. (not to mention stuff like snuff laws, which don't). It's not a given.
So, would you support such a law, if it fixed the strict scrutiny issues? If so, I apologize for doubting.
parody and satire tends to be considered protected speech. You can argue it shouldn’t be, but you would have a steep hill to climb to convince a court.
Eh, I think you can make a pretty reasonable argument that this specific type is in fact 'truly harmful', even if satire/parody tends to be protected in general. The fact that it's satire/parody doesn't remove the fact that's essentially still NCII. It's a bit borderline, but it's not crazy.
So if you think this law was designed to prevent this kind of parody, then you are admitting it is unconstitutional.
I don't think we really have precedent on this, yet. It'd likely be ruled unconstitutional based on past precedent, and it's a steep hill. Even if you require it to fit Stevens, you could fit it into obscenity (or less likely, integral to crime), if you squinted hard enough.
I'd also qualify it as "currently unconstitutional", rather than just "unconstitutional". Bit of a nitpick, but just calling it unconstitutional implies it is unconstitutional in the abstract and can never climb that hill. That's a distinction we're usually careful about when talking about stuff that may be blessed by SCOTUS in the wrong direction (see: Paxton).
Let’s start with the basics: apps (mostly powered by various AI tools) that are used to produce modified imagery, especially stripping people of their clothes are… bad. They should be socially shunned. People using them to objectify or sexualize others are doing bad things, and people should judge those who use those apps accordingly. This is not a defense of those apps.
Socially shunned is not legally banned. This law may not be perfectly written, but if you are not willing to ban them with a well written and targeted law, you are in fact defending those apps with some very carefully chosen wording.
It's amazing how Musk always manages to find the most problematic parts of a Techdirt stance, and highlight it, by being the biggest asshole possible.
You also left out two pretty important and interesting parts of Kathryn's thread, and the law. 230, and strict liability.
And to be clear, before I get strawmanned:
Of course, no one wants to hear that the law is badly drafted. Lots of people want to ban nudify apps
I'm happy to read about how the law is badly drafted. It has real problems. It also has parts people like Techdirt won't like, that aren't problems. The issue isn't the former, but the former shouldn't be used to smuggle in the latter.
then what is the argument about distillation doing on the list at all? Testing catches dangerous capabilities regardless of how the model got them.
Because that assumes you can subject every model to testing, or do anything in response to a test. Something you can't guarantee. Let's say we all broadly agree all models should be tested. Then what? You can't compel e.g. China to test. Or even go a step further, lets say one of those models 'fails' the test, then what? You can't put the genie back in the bottle.
Of course, if you're concerned with distillation that's kind of fundamentally incompatible with having a consumer facing product in the first place, which is kind of a problem for Anthropic.
Now, to be clear, as with any technology, you can claim that a more open, more widely available, more powerful version can be misused. But that has always been the case and we, in the US, have tended to default to allowing the technology to proceed, and figuring out ways to minimize the dangers/increase the good uses, rather than resorting to assuming the tech will be abused and working backwards to block all possible abuses.
TBF, most of Ant are in fact true believers in the whole existential threat issue with AI, and their stance needs to be interpreted through that lens to make any sense. (And to make it more fun, the fixes for that are basically diametrically opposite concerns about authoritarianism/market capture). They don't often approach it directly because they know it's not mainstream opinion, but it's not hard to square their stance once you factor that in.
Even if there’s some truth to it, it’s awfully convenient that the same argument also happens to justify banning, punishing, or limiting the cheaper, more open, more user-controllable alternatives.
I mean, it's intentional. Those marketing stunts are basically them trying to give evidence of how it can go wrong, by showing a toy version of “weaponizing pandemic-level viruses.” without actually making a pandemic-level virus, before it gets there.
Yeah, strictly legally I think Google is going to have a hard time for exactly that reason. But if we're talking normatively about "pulling up the ladder" rather than just descriptively about the law, I think we have to be a bit careful. There can still be concerns even if it's not well covered by copyright law
Google’s entire business was built on scraping the web. Suing someone else for scraping Google sure feels like pulling up the open internet ladder up after themselves.
I don't think it's really fair to conflate scraping the web (something SerpApi can still do itself just fine) with scraping Google, which is essentially just piggybacking on Google's work. The ladder's still there, SerpApi just wants the shortcut.
Google tried to argue that it somehow has the support of copyright holders to protect their work with SearchGuard, but the court is not impressed.
I'm not sure how I feel about that. If Google is scraping my data to use in fair use (something I don't get to control), I do kind of expect Google to also be a good custodian of that data. I don't really think I should be expected to have to babysit how Google does that.
That section is premised on the idea that the author has some sort of agreement/control, but for something like scraping, the author hasn't been consulted. (And obviously, the wording has to be a bit careful, because it would silly if a third party could put up a restriction against the author's wishes).
When the argument is only selectively used for some speech laws, to argue those laws are bad/unworkable because of it, but doesn't get mentioned otherwise (for say, these election laws, which is having the same bad-guys-with-pens problem), yes that is in fact the logical conclusion. Seems like pretty fair game to bring it up, when a counterexample happens to come along.
This isn’t a federal investigation, it’s a local one
I'm aware. That doesn't change the actual points I'm making.
While it is entirely possible that the sitting DA will choose not to prosecute the case for partisan reasons, I’m guessing that’s not who you’re talking about as people don’t typically use the word “administration” to describe a district attorney’s office.
What I'm talking about is, that both the local DA and the federal government will have ways to make this a problem for Musk. This isn't a strictly local issue only the DA can touch.
Given that the DA is likely not to prosecute for partisan reasons (and assuming the seat doesn't get flipped any time soon. And even if it does get flipped, you have other hurdles like state courts that have already done things like decline to stop the original payments), that means your likely avenue for recourse in the near future is probably going to be via the future federal government. You do not have to pin all your hopes on the unlikelihood of the DA doing something.
This does not mean the sitting DA (or this particular investigation) is a part of the federal government/administration. There just happen to be a lot of bad guys with pens in the way of a local prosecution, and it's likely not going to be enforced despite being a clear violation of the law. The federal government is likely going to be the available lever to be pulled, if you actually want to enforce consequences for breaking the law.
Which is a roundabout way of saying it kinda seems like you didn’t read the article and just jumped down to the comments to tilt at strawmen.
I did in fact read the article. And you can tell, since the two points I'm making only makes sense in response to the content of the article. I mean, heck, one is literally a direct quote from this article. Do you think I grabbed it with magic?
I don't think it's a strawman to point out an inconsistency in which bad guys holding the pen supposedly make particular laws untenable or not. Particularly when it directly ties into wanting the law to be enforced and how to make that happen, something which is a big part of the article. There's some pretty clear through lines.
Turns out So here’s the only test that matters before you back a law like this: imagine the politician you distrust the most holding the pen.
doesn't apply just to speech laws.
At some point, those who break the law with impunity have to be held accountable for their actions.
The bad guys are holding the pen. If you want the law to be enforced, it's going to have to come from making the next administration do it.
I find this article hard to parse, and it seems a bit muddied about what exact issue is. At times it's written as if the specific implementation/uncertainty is the issue, and other points it seems to be against the FTC being involved at all. In particular:
When DoNotPay promoted a “robot lawyer” as comparable to a human professional, then-Commissioner Ferguson rightly voted to hold it accountable. When Workado exaggerated the accuracy of its AI-detection product, the FTC, with Ferguson as chair, ordered it to stop making unsubstantiated claims.
I don't really see how these two "good" examples are differentiated from the First Amendment concerns it brings up elsewhere? So with that, I can only really touch on a few points:
Only in this incarnation, she pulls a novel theory of deception from her regulatory carpetbag to control what AI chatbots say.
I wouldn't really call this a novel theory. This seems pretty straightforwardly a deception. But putting that aside, it's wrong to say this controls what AI chatbots say- Having to disclose something doesn't control what AI chatbots say. You can still say the thing.
Consumers might struggle to reconcile the chatbot’s biased-but-true disclaimers
If consumers are struggling with that, that does seem to suggest it is in fact deceptive. But regardless, this seems fairly easily squared under a typical 'reasonable person' standard.
The design choices underpinning large language models make them legally indistinguishable from video games and social media.
Mechanically, there is a major difference that is being missed here. With an AI system, you can have disclosure of e.g. your training set, your fine-tuning, etc. And as a bonus, requiring that disclosure is all completely content-neutral. That's not really possible with subjective editorial judgement.
I can see some reasonable concerns about the exact wording used, and concerns about e.g. subjective standards to suppress disfavored speech (particularly given this administration), but the article seems to bounce between that concern and not wanting the FTC to be involved regardless of implementation, in a way that isn't really clear to follow.
and users at no point needed to switch to a brand new app, continuing instead to use the very app we were told was a security nightmare.
This is a bit disingenuous.
Someone else having control of an app does actually change the security implications, in terms of where the data can be accessed, potential future vulnerabilities introduced, etc. And that's ignoring any updates to the app, even if it's the "same" app. It's not perfect, since you never know if you missed something, but it is pretty meaningfully different from a security perspective.
The real lesson here appears to be that the earlier concerns were exaggerated.
The lesson is that GOP/Trump believing those concerns were exaggerated. That doesn't mean the underlying issues themselves didn't exist, because as we all know, Trump/GOP will happily sell the country out for a penny. See any of a million decisions with respect to e.g. Russia, JCPOA , the bribe plane, Saudi nukes, etc.
Narcissistic grifter continues grifting doesn't tell you anything about the underlying issue, it tells you they're equal opportunity grifters.
A growing body of research suggests that one-size-fits-all solutions are not the answer
I don't think anyone thinks one-size-fits-all solutions are optimal. The trade off is supposed to be that it's simpler to message/implement, at the cost of being less optimal.
I guess my question would be, how do you actually get parents to do this? Compliance with one-size-fits-all is already pretty low. Because otherwise this kind of just ends up back at thoughtful and engaged parents are good.
I kinda think that the onus should be on the people demanding that we “do something” to give us their proposals, rather than demanding the people who say “those are bad proposals and make things worse” give people who want to do something proposals for what they should do.
I mean- they did, didn't they? And they're winning, in terms of public opinion. So the question becomes, how do we adjust to that? Do you keep making the same argument and hope something changes? I think that stance was maybe justifiable 10 years ago, before the dam broke. Not so much anymore.
Do you stick to “those are bad proposals and make things worse”, or do you pivot some of that effort towards "these are bad proposals, but here is a way we can mitigate the harm and make it less bad"?
To be clear, I don't think you should give up saying “those are bad proposals and make things worse” entirely, because they're not mutually exclusive. But I think a lot of the OG privacy crowd gets fixated on “those are bad proposals and make things worse” being the correct answer to the exclusion of opportunities to get us to less bad. And the net result is things end up slightly worse than they could be.
And even when “those are bad proposals and make things worse” is on the table, they're kind of out of touch in pitching it, because they pitch it like it's 10-20 years ago or like they're talking to each other.
Like how would that make sense in any other context?
The way I look at it is, what is likely to minimize the most harm. And I think (conditional on the fact that for now you've lost the argument to just not do it) spending some time on mitigating harm on laws that are going to be passed anyway, will mitigate more harm than just sticking with saying it's bad.
If we were winning, it'd be another story. Clearly, do nothing is better than bad proposal and less bad proposal. But conditional on being stuck with picking between bad vs less bad, it is worth putting in some effort to end up at less bad, because we're getting one of them.
To use a silly random analogy: lets say you know your kids are going to go drinking at parties. You can tell them not to do that, but you know they're going to do it anyway, because they're kids. Given that, you should make it clear you are ok being a designated driver. Not just keep telling them to not drink. Is the correct answer not to drink? Yes. But they're not going to do that, so the best you can do is mitigate the risk by being a DD.
Being correct hasn’t helped and has hurt? Fuck that. That is wrong and stupid.
To be clear, I'm not saying being correct hasn't helped. I'm saying the way you're trying to convince people you're correct is no longer effective as it used to be, and you aren't adjusting to that. And that you're leaving material gains on the table because of it.
Being correct and convincing people you're correct are two different things. Being correct while sidelined off in a corner by yourself doesn't mean anything, when it comes to politics. You also have to sell it to the public. The issue is not the correctness.
These efforts have been a disaster, just as we predicted.
And yet, they're spreading. Clearly we're not convincing people these efforts have been a disaster.
Maybe at some point there will still be a backlash, but it hasn't manifested yet. I think it is foolish to pin all our hopes on that when it isn't in sight.
You would prefer that I lie to make some politicians feel better? Because I won’t do that.
I would prefer you tell the truth. Saying kids will always be able to bypass it isn't true- the reason we find state coercion so concerning is because you can't guarantee winning against the state in the long run, if it wants to be coercive. That's why we have so many protections against the state.
You can now, and likely will for awhile, bypass it. But it's never a future guarantee (and even if someone technologically savvy can, it's still leaving the less savvy masses behind). If you knew you could always bypass it, it frankly wouldn't matter. But the reality is, eventually those loopholes can be tightened up. That's why it's scary.
But more than just honesty, and the reason I harp on it, is I think it actively undermines the effort to combat them. People are going to be less concerned if you tell them there will always be workarounds for their own use. The reality is, people care more when they think it will directly affect them. This isn't about making the politicians feel better- it's about not undermining the other message about how what they're doing is dangerous when going for the laughs. That doesn't mean you can't laugh at the laws for sucking. Just that it's better to avoid promises like that there will always be a workaround, because that undersells the threat.
I’ve long argued that users themselves are the ones who should be liable for any violative actions they make rather than the platform but I don’t recall ever seeing a platform actually sue one of its own users for using their own tools.
I mean, it's the logical conclusion of that stance. If you truly believe only the user should be liable, why wouldn't indemnification apply? It's just transferring the liability where it supposedly should go- the user.
Which all sounds well and good if you simply ignore the whole “yeah, but your CEO kinda laughed about and promoted this use of the tool” part.
We were told the law should ignore that part. After all, that doesn't make it first party speech.
The whole point of this law is it introduces liability for the platform for content a user produces. If that's your stance, you're saying you still wouldn't support a version of this law that passes strict scrutiny. That is a defensible stance, but is explicitly not the “if this law just fixed strict scrutiny issues, I’d support it" stance that is being implied. It's the other one.
Addendum: I also forgot to mention that a law that satisfies strict scrutiny doesn't necessarily satidfy your long time stance of users themselves are the ones who should be liable for any violative actions they make rather than the platform, either
Yeah, strictly legally I think Google is going to have a hard time for exactly that reason. But if we're talking normatively about "pulling up the ladder" rather than just descriptively about the law, I think we have to be a bit careful. There can still be concerns even if it's not well covered by copyright law
Turns out So here’s the only test that matters before you back a law like this: imagine the politician you distrust the most holding the pen. doesn't apply just to speech laws. At some point, those who break the law with impunity have to be held accountable for their actions. The bad guys are holding the pen. If you want the law to be enforced, it's going to have to come from making the next administration do it.
I find this article hard to parse, and it seems a bit muddied about what exact issue is. At times it's written as if the specific implementation/uncertainty is the issue, and other points it seems to be against the FTC being involved at all. In particular:
I don't really see how these two "good" examples are differentiated from the First Amendment concerns it brings up elsewhere? So with that, I can only really touch on a few points: I wouldn't really call this a novel theory. This seems pretty straightforwardly a deception. But putting that aside, it's wrong to say this controls what AI chatbots say- Having to disclose something doesn't control what AI chatbots say. You can still say the thing. If consumers are struggling with that, that does seem to suggest it is in fact deceptive. But regardless, this seems fairly easily squared under a typical 'reasonable person' standard. Mechanically, there is a major difference that is being missed here. With an AI system, you can have disclosure of e.g. your training set, your fine-tuning, etc. And as a bonus, requiring that disclosure is all completely content-neutral. That's not really possible with subjective editorial judgement. I can see some reasonable concerns about the exact wording used, and concerns about e.g. subjective standards to suppress disfavored speech (particularly given this administration), but the article seems to bounce between that concern and not wanting the FTC to be involved regardless of implementation, in a way that isn't really clear to follow.