←back to thread

Anthropic’s paper smells like bullshit

(djnn.sh)
1160 points vxvxvx | 3 comments | 16 Nov 25 11:32 UTC | HN request time: 0.729s | source

Earlier thread: Disrupting the first reported AI-orchestrated cyber espionage campaign - https://news.ycombinator.com/item?id=45918638 - Nov 2025 (281 comments)
Show context
gadsnprch ◴[16 Nov 25 18:01 UTC] No.45947024[source]▶
Why isn’t Anthropic held liable for crimes committed with their product? I feel totally befuddled as to why that is not the conversation, but rather Anthropic is doing a victory lap like they are the good guys despite their product enabling widespread fraud while they amass outrageous, undeserved, profits. Why is Anthropic not liable?
replies(1): >>45948781 #
saagarjha ◴[16 Nov 25 21:55 UTC] No.45948781[source]▶
Because deciding how much culpability they have is not a solved problem.
replies(1): >>45949528 #
1. gadsnprch ◴[16 Nov 25 23:41 UTC] No.45949528[source]▶
Thanks for responding. Solving it will involve public discourse. The negative externality will not be forever ignored and frantic, knee-jerk, legislative or judicial solutions are rarely optimal. Everyone, including Anthropic, benefits from starting the culpability discussion now, ideally in a context just like this. Maybe that’s exactly what Anthropic is doing by framing this news as “we stopped some cybercrime” rather than “we were involved in some cybercrime.” But smart people shouldn’t fall for such a blatant shifting of corporate liability onto the public, imo, and that’s why I’m confused. I must be missing something fundamental.
replies(1): >>45963955 #
2. saagarjha ◴[18 Nov 25 11:49 UTC] No.45963955[source]▶
What kind of solution do you foresee here?
replies(1): >>45967937 #
3. gadsnprch ◴[18 Nov 25 15:56 UTC] No.45967937[source]▶
Well, if it were not for the fact that our two party system is controlled by donors and lobbyists I would hope for a legislative solution. If your product is used to defraud people, you have to reimburse the victims seems like a reasonable sort of law to enact. Such legislation would encourage companies to do more than give vague lip service to “guardrails.” I’ll admit that I’m not an anarchocapitalist who thinks we need to put all our resources, and the resources of future generations via overspending, into building ever more impactful AI, so maybe I don’t belong here. The ai worshippers have been making claims for several years now about ai curing cancer and solving all our problems: where’s the beef?