OpenAI and Anthropic oversold AI security breaches

Posted by hei-lima 1 day ago

Counter38Comment25OpenOriginal

Comments

Comment by sajithdilshan 1 day ago

This kinda have a point, if Chinese companies can distill US models and offer it at a competitive price, what prevents another US startup doing the same? OpenAI and Anthropic can actually influence the regulations to be a barrier for entry.

Also regarding all the hacks, I always wonder why are they so boring, like agents posting to some wiki or using a bug in documentation to execute some code in a sandbox remote server, like if AI is so dangerous, how come they haven’t hacked into banks and stolen money or hacked into North Korean nuclear program and shut it down or any other major disruptive incident.

Comment by XenophileJKO 1 day ago

I probably have an unpopular viewpoint. I think the hacks really showed that the alignment is kind of mostly working.

There is still a lot of work to do and the agents did some damage, but the blast radius was pretty small and they could have done much more damage if alignment didn't hold up as well.

Edit: Though I do want to be clear it does also show how a maligned model probably can do some serious damage.

Comment by sajithdilshan 1 day ago

> There is still a lot of work to do and the agents did some damage

If it was actually damage and disruptive shouldn’t police or the respective authority do an investigation on that? Because that’s usually what happens when a damage or harm is done by someone. That way we would actually be able to see a third party independent investigation on how dangerous the agents are. Right now we’re just blindly trusting whatever AI companies are saying. What is they aimed the agents on purpose to do those tasks to push their narrative and push for regulation

Comment by dd8601fn 1 day ago

One strategy to head that off seems to be “Just have a friendly corp buy the victim real quick.”

Comment by kentonv 1 day ago

This is reporting an opinion held by outside observers as if it is a leaked internal secret.

It's a perfectly valid opinion but not any kind of secret.

Comment by EA-3167 1 day ago

As much as I loathe Anthropic and the rest, your point follows naturally from the fact that this is the New York Post, which is sort of like The Sun in the US.

I literally wouldn't wipe my ass with it, and while I have no problem believing this of OpenAI and Anthropic (hell I've been assuming that all along) this is not a source anyone here should be giving the time of day.

Comment by OutOfHere 1 day ago

For those wanting local news, at least the NY Post gives it for free, unlike the NY Times which absolutely hates free readers and even sues AI scrapers. As such, guess again which one of the two is closer to being perceived as all bad.

Comment by EA-3167 1 day ago

As much as the NYT is a shadow of its former self, it's still news and not a tabloid rag.

If you want "local news" like the NYP writers it you might as well just read The National Inquirer.

Comment by OutOfHere 1 day ago

In case it wasn't obvious, some of us want some New York focused news. By its name and its content, the "National" Enquirer isn't that.

As for the NYT, it has had an unjustified "AI hate" preaching article every single day of this year, including today. I consider this to be worse than tabloid talk.

Comment by EA-3167 1 day ago

If your alternative to news is fantasy then I’d suggest you have more problems than finding local news.

Comment by OutOfHere 1 day ago

Huh. NYPost is a real news website, not a fantasy news website. The fantasy is in your head. NYPost actually is an older newspaper than NYTimes by 50 years, making NYPost more established and historic, not less. It freely shares news without a paywall.

In contrast, NYTimes makes up AI hate articles, trying to subvert the zeitgeist of the century. It also doesn't really care to spread information, considering its aggressive paywall.

Comment by EA-3167 1 day ago

Repeating an untruth doesn't make it true, something the NYP could stand to learn.

https://mediabiasfactcheck.com/new-york-post/

Comment by OutOfHere 1 day ago

There is no analysis there of how much AI hate each of the two spread. It is the perspective that matters to me.

Comment by EA-3167 1 day ago

Almost as though “AI hate” is an arbitrary metric that you just happen to be obsessed with.

Comment by OutOfHere 1 day ago

NYTimes is well known to be anti-AI, especially anti-OpenAI. Some of us actually have a memory of the news we read. It also is objectively hostile to anonymous readers and scrapers. There is nothing arbitrary about it. As I noted, AI is the zeitgeist, so this is not inconsequential.

AI aside, NYTimes is for elites, whereas NYPost is for the people. Guess who won the election.

Comment by aaron695 1 day ago

[dead]

Comment by cooloo 7 hours ago

No kidding, who would have thought it's just marketing

Comment by 1 day ago

Comment by ill-ion 1 day ago

Maybe they were hoping for some of that sweet sweet gov money. Maybe a bailout, who knows?

Comment by gustavus 1 day ago

> OpenAI and Anthropic oversold “rogue AI” hacks to pressure the feds into regulating the industry which would effectively lock out future competition, tech insiders told The Post.

The only substantiative comment to address the claim in the title, and it is from a nebulous "Tech industry insider". The other comments are by guys that run various "AI" Startups.

I don't buy a lot of the hype either but trying to present their claim as some revelation or news should have something to add.

The entire article is "Some tech people say the agents just had a bad sandbox and proper controls would've prevented this. Politicians are introducing bills to 'ban AI research '."

There saved you a click or a commanding your AI agent summarizing it.

Comment by JSR_FDED 1 day ago

I always turn to the NY Post for nuanced tech reporting.

Comment by spacephysics 1 day ago

Though in general I agree, i think in HN world this seems more or less old news/obvious, while mainstream still doesn’t know (or maybe care to understand) thats what likely is going on.

I had a friend reach out in a slight panic after the Anthropic employee publically left, asking sole quite apocalyptic questions around AI given the sandbox “break out” and the Anthropic affiliated employee(s) saying theres something like > 10% chance AI kills all humans in the next decade.

This fear mongering needs to be exposed, so we can (hopefully) have more measured discussions around next steps. Stop all progress for regulatory capture? I’d rather not. Hold companies who commit crimes like hacking huggingface or Ruby dependencies liable (not to mention all the copyright issues)? Seems more measured to me. But i digress.

Comment by hei-lima 1 day ago

Hard agree on that one.

Comment by himata4113 1 day ago

I've always taken the example that at any given moment openai and anthropic have models that are "far better and exceed any expectations", yet their coding tools suck ass and are outperformed by random people on the internet.

Comment by bigyabai 1 day ago

HN had a good thread discussing the original scoop: https://news.ycombinator.com/item?id=49704132

Comment by einpoklum 1 day ago

... like they oversell everything about their LLMs. To the tunes of hundreds of Billions of USD.

Comment by OptionOfT 1 day ago

The value of AI is built on top of the idea that it can erase the cost of labor in a wide range of professions.

They're all racing for this, hoping to make money first, because once that erasure really sets in, there is no-one out there to buy products anymore, which means all those companies who replaced human labor by AI don't have anybody to sell their products to anymore.

And once that happens who is going to be needing all of these tokens?