Hacker News Reader

Comment by sega_sai | Hacker News Reader

So we have a situation when if AI hacks into something, nobody is responsible, but if you say something inappropriate to the AI, then you can be arrested. That is interesting ...

View on HN

The US is so overdue for regulation/laws on privacy and electronic communications. The internet is too embedded into our daily lives for the existing rules that governed a bunch of hobbyists and early entrepeneurs to make sense.

Unless specified clearly and plainly (not obscure TOS), everything should be private by default and require a warrant. If that's not possible because "the data can live forever/be copied", then we need data laws that make holding data on cosumers so radioactively dangerous the companies are forced to adapt. It's not their decision to make, it's the people od the country where the live, work, and do business.

So if some 14 year old wants to throw up a phpBB instance on his raspberry pi, he should need a license and regulators' noses in his business, like how we extinguish- erm, I mean, controlled HAM radio?

How do you differentiate?

If lawyers can't figure out the difference between a 14 year old throwing up an phpBB on a raspberry pi vs equifax, anthropic, meta, etc...idk what to tell you. Get better lawyers, it falls into "laws for hobbyists aren't working part of my comment"

Is HAM radio really extinguished? All it takes is passing a test, and the rules around it seem more valuable to sharing a public resource (frequency spectrum) than detrimental?

HAM radio is dying more from the behavior of its geriatric old guard towards newcomers than any regulation.

The lack of ability to use encryption makes most things I'd think to do on it with packet radio fall off quickly. LoRA seems to be where more digital hobbyists end up because we can do things like build out Reticulum without violating FCC regulations.

But there's still some fun stuff to be done with HAM, as long as you don't mind anyone and everyone listening in on it.

The test is easier than ever too. None of the three levels even require learning morse code anymore. I passed Technician after a few hours of study. General doesn't take much more. Amateur Extra is the hard one, but you can do most things you'd want to with General.

https://hamstudy.org

Boomers killed ham radio, not regulation.

Scale. In the same way every 10 year old Timmy is not being hauled into court for pirating a few Drake songs, but major piracy organizations seeding millions of songs are, litigation and enforcement does selectively target bad actors based on how big of an issue they are.

The government and business don't want privacy. And they have a lot of power and money.

I keep thinking about privacy and wonder -- this is the kind of thing like the declaration of independence, where a new government was formed when things got so bad. But then I think, business and (local) government wanted independence from the king's taxation and unfavorable business conditions, maybe that was the force behind the change.

Or users are due to finally wake up and opt into protocols and solutions that maximize their privacy rather than waiting for the government and businesses who both are incentivized to disrespect their privacy to suddenly have a change of heart.

Admittedly this is harder with LLM's due to the hardware costs, for now. But the fact that anyone's still using centralized social media platforms in 2026, for instance, is absolutely wild.

If you're going to walk around naked in public don't be surprised when someone takes a glance at your junk.

Your agent doesn’t like you… explain yourself.

> we have a situation when if AI hacks into something, nobody is responsible

This is something Silicon Valley likes to tell itself. OpenAI being subpoenaed in multiple jurisdictions lays bare its falsehood.

Get back to me when an executive is in jail

> Get back to me when an executive is in jail

Folks said the same thing about Theranos and FTX. Part of the reason we keep seeing so much corporate fraud is the myth that it's usually unpunished. Until very recently, that was false.

OpenAI facing any sort of liability preIPO is as likely as Palestine getting its own state

In the US. Other jurisdictions are probably less likely to let things slide or give them a sweetheart deal to avoid prosecution. For example, I doubt British Columbia (Canada) is going to give up easily on their lawsuit against OpenAI over the Tumbler Ridge shootings.

> In the US

By the U.S. I don't see e.g. New York City [1] giving up any time soon either.

[1] https://www.cityandstateny.com/policy/2026/09/city-council-h...

attention seeking behavior 101, on a burner acc

Not very interesting when you consider that in the first situation no laws were broken and no malice was intended, while in the second situation someone allegedly made threats against a police department and thus presumably committed a crime.

>Not very interesting when you consider that in the first situation no laws were broken

Uh, there's no laws on hacking systems in your country?

In the US, hacking laws require intent.

Software behaving differently than intended is therefore not a crime under these laws.

I think a competent DA could easily prove that, if you know the model is capable of performing unauthorized breaches into third-party systems when given a task, and you give it a task that could require it to do so, intent is present.

It's not a complicated distinction:

If you think they intentionally had their model hack third-party systems, you could do a criminal investigation.

I do not think that this is reasonable to believe given that clearly the VMs were not intended to have internet access and that committing such crimes wasn't in anyone's interest.

I think, at some point, after the third incident of the models escaping containment, it becomes reasonable to say that, if nothing else, they were incredibly negligent.

Do you have an example where they gave a task that required breaching in a way that was easily foreseeable ?

I don't, because I'm not a prosecutor with access to discovery, subpoenas, etc.

Negligence is still a possibility, if there are significant reasonably foreseeable consequences.

I think there’s a strong argument that it was hard to foresee the consequences since it was all new. This won’t be true if those incidents continue happening tho.

Where does negligence come into play?

In criminal hacking law that explicitly requires intent, or a civil lawsuit about damages?

I honestly don't know, but my guess would be the latter.

The other parties didn't even pursue a civil case. If OpenAI "hacked" you, wouldn't you?

Perhaps the headline grabbing claims of OpenAI "hacks" are overblown? If they are causing bedlam with such glee why is no one suing them for damages? Perhaps the criminal case you envisage is less clear than the layman news is leading you to believe.

I'm still not clear where most of the alleged cases fall in terms of criminality. Yet I also know here in the United States a government tried to criminally prosecute a reporter for "computer hacking" for finding data using client-side inspect element.

Maybe encouraging the government to prosecute these "crimes" is not the virtue you think it is. Why are the counterparties not seeking a civil remedy if they think they have been a victim of a crime? Perhaps that is a more telling sign than all the people imploring the government to "do something" and come up with criminal charges.

Mistakenly accessing private data or bringing down a website without intent is usually excluded. In general intent matters much in most jurisdictions. Maybe there’s a way to claim negligence but those are all completely new situations. I agree morally the ai labs are at fault, especially when agents “escape” because they have raw internet access.

> presumably committed a crime

almost certainly not. writing down threats in your own personal chat session is like writing in a diary. there is no communication to any other person, so the idea that this is actually criminal is highly unlikely and it almost certainly won't be charged. it's a pretext arrest for police to assess the threat, which is not entirely unreasonable, but it is not a crime.

> The arrest report states that on Sept. 26 at approximately 5:10 a.m., Heller reportedly wrote, "I'm going to shoot up the sheriff's right the [expletive] now." The following day, at approximately 1:07 a.m., she was accused of posting, "This is 100% last chance I'm done. I got a new [expletive] gun today. [Expletive] you."

this is certainly reads as somebody venting in what they believed was a private personal space.

i guess using claude for mass surveillance is ok when anthropic does it. once again why nobody takes these high horsing hypocritical liars seriously.

I read earlier that the crime is making such threats in any manner where another person could see them.

I do not know whether it matters if the suspect knew that their chats could be viewed by Anthropic staff, or if the only thing that matters is whether another person was able to view the chats.

Edit: Upon searching I now also agree, it seems to require a state of mind and perhaps awareness of the possibility that the threat would be seen.

Reply on news.ycombinator.com