141–150 of 593 posts
I was hoping for a moment that this meant they had come up with a design that was safe against lethal trifecta / prompt injection attacks, maybe by running everything in a tight sandbox and shutting down any exfiltration vectors that could be used by a malicious prompt attack to steal data. Sadly they haven't completely solved that yet. Instead their help page at https://support.claude.com/en/articles/13364135-using-…
This looks useful for people not using Claude Code, but I do think that the desktop example in the video could be a bit misleading (particularly for non-developers) - Claude is definitely not taking screenshots of that desktop & organizing, it's using normal file management cli tools. The reason seems a bit obvious - it's much easier to read file names, types, etc. via an "ls" than try to infer via an image. But it a…
Earlier quoted context omitted.
Have you ever used any Anthropic AI product? You cannot literally do anything without big permissions, warnings, or annoying always-on popup warning you about safety.
Claude code has a YOLO mode, and from what I've seen a lot of heavy users, use it. Fundamentally any security mechanism which relies on users to read and intelligently respond to approval prompts is doomed to fail over time, even if the prompts are well designed. Approval fatigue will kick in and people will just start either clicking through without reading, or prefer systems that let them disable the warnings (just…
This looks useful for people not using Claude Code, but I do think that the desktop example in the video could be a bit misleading (particularly for non-developers) - Claude is definitely not taking screenshots of that desktop & organizing, it's using normal file management cli tools. The reason seems a bit obvious - it's much easier to read file names, types, etc. via an "ls" than try to infer via an image. But it a…
> Claude is definitely not taking screenshots of that desktop & organizing, it's using normal file management cli tools Are you sure about that? Try "claude --chrome" with the CLI tool and watch what it does in the web browser. It takes screenshots all the time to feed back into the multimodal vision and help it navigate. It can look at the HTML or the JavaScript but Claude seems to find it "easier" to take a screens…
And I do know there are ways to hide data like watermarks in images but I do not know if that would be able to poison an AI.
Earlier quoted context omitted.
> It's just that the user fundamentally cannot provide informed consent The user is an adult. They are capable of consenting to whatever they want, no matter how irrational it may look to you.
Uh, yes? What does that refute?
In any context, I really dislike software that prevents me from doing something dangerous in order to "protect" me. That's how we get iOS.
The user is an adult, they can consent to this if they want to. If Anthropic is using dark patterns to trick them that's a different story--that wouldn't be informed consent--but I don't think that's happening here?
Earlier quoted context omitted.
> Why are you marketing a product clearly for business use Huh? Their "individual" plans are clearly for personal use.
Is that why you can enter a business id on the payment form? Just read the marketing page [0]. The whole thing is aimed at people running a business or operating within one. [0] https://claude.com/pricing/max
I was hoping for a moment that this meant they had come up with a design that was safe against lethal trifecta / prompt injection attacks, maybe by running everything in a tight sandbox and shutting down any exfiltration vectors that could be used by a malicious prompt attack to steal data. Sadly they haven't completely solved that yet. Instead their help page at https://support.claude.com/en/articles/13364135-using-…
There is much more to do - and our docs reflect how early this is - but we're investing in making progress towards something that's "safe".
Isn't this just a UI over Claude Code? For most people, using the terminal means you could switch to many different coding CLIs and not be locked into just Claude.
Most people have no idea what a terminal is.
I was hoping for a moment that this meant they had come up with a design that was safe against lethal trifecta / prompt injection attacks, maybe by running everything in a tight sandbox and shutting down any exfiltration vectors that could be used by a malicious prompt attack to steal data. Sadly they haven't completely solved that yet. Instead their help page at https://support.claude.com/en/articles/13364135-using-…
> (I don't think it's fair to ask non-technical users to look out for "suspicious actions that may indicate prompt injection" personally!) It's the "don't click on suspicious links" of the LLM world and will be just as effective. It's the system they built that should prevent those being harmful, in both cases.
Not only is the attack surface huge, but it also doesn't trigger your natural "this is a virus" defense that normally activates when you download an executable.