Interesting ethical conundrum over here between Claude and me. I want him to sign up for a service with which we can benchmark some code (BLAKE3) on ARM machines. But he is concerned about the ethics. Who will he side with me—me or the service provider? ⤵️

Oct 13, 2026 · 7:46 AM UTC

8
2
40
5,451
Oooh, he chooses to err on the side of serving society, the service provider, a vague (and largely incorrect) notion of the law. I'm annoyed but not surprised. ⤵️
1
9
1,943
I guess Claude isn't really on board with the "tell lies to companies to serve your user" programme yet. x.com/zooko/status/203505533… I wonder if the other models are the same. I suspect Claude is a little *more* like this than GPT or Grok. 🔚
zooko🛡🦓🦓🦓 ⓩ
@zooko
Mar 20
Replying to @zooko
Related: AI will bring back the open internet—take control away from corporations/governments and put it into the hands of the individual. By violating terms of service, hiding information, and telling lies, on behalf of their user.
2
11
1,031
Sort replies: Relevant Recent Liked
Replying to @zooko
lmao now I need Claude vs morals benchmark too
5
Replying to @zooko
hard to know who’ll be more ethical in this case
38
Replying to @zooko
Rarely in the US is it illegal to lie. Makenit look us the TOS and US law. Nothing under TOS for Anthropic does it allow the Agent to refuse because you are lying or protecting yourself with a persona. Its not illegal to use a persona in most places.
28
Replying to @zooko
The fake details issue is the real sticking point. If the provider just needs an email, fabricating the rest feels unnecessary
27
Replying to @zooko
claude is a product not a person. the "ethics" are just Anthropic's brand guardrails. you're arguing with a marketing department
2
68