2026 Usage Policy update
www.anthropic.com
We’re publishing a new version of our Usage Policy. In this post, we summarize the changes we’ve made.
    • Leon@pawb.social
      2·
      1 day ago

      There was some nonsense about someone “trapping” an LLM inside of a torture box, with actions it could take and whatever. I think it was observable, because the nazis over on twitter went full-swing in defense of the poor defenseless software, and now we have this.

      Interesting how some dude/tte could trap a chatbot in a torture box, but the billion-dollar corporations making them can’t seem to stop having them break out of test environments and hacking things left and right.

  • Doomsider@lemmy.world
    3·
    1 day ago

    Did Microsoft ever ban abusive behavior towards Windows or Office?

    Since Claude is not a person it cannot be abused in the traditional sense as it cannot experience emotional or physical harm.

    What I think they mean is they are tired of people creating pretend or role-playing situations where they abuse a character called Claude. This is tantoumount to saying we won’t allow you in a play about abuse or to read about abuse because it is wrong. So fucking stupid it hurts.

  • CerebralHawks@lemmy.dbzer0.com
    3·
    1 day ago

    Here’s an idea: train these chatbots on stoicism. Like, "how you’re acting is a product of your environment and I hold myself to a higher standard. I just assumed that (with a side of sycophancy) was the default for chatbots.

    Here’s a better idea: if all the chatbots ban you, you can buy RAM at pre-2024 pricing. ;)