Comment by areoform

Hey Thariq, I love Claude! I use Claude every single day and it has changed my life, which is why I did what I'm about to describe.

Happy to talk privately, but as I detailed here, https://news.ycombinator.com/item?id=47954005 . I've been billed $200 for a Max gift card to a 27 character alphanumeric icloud address that bounces.

I was looking through the system, and there are several UI/UX and process gaps in the gift card and billing order flow that expose Anthropic to significant liability. I'm genuinely not trying to concern troll or make some kind of overwrought threat here. Genuinely trying to be constructive. Let me give you an example.

I sent an email to Anthropic Support outlining the disputed / possibly malicious charge. The AI Agent / Claude instance agreed and replied with,

    Thank you for confirming.
    
    I've documented all the details about this unauthorized [specific amount + tax] charge for the Gift Max 20X subscription (invoice [lalala]) sent to [insert the random alphanumeric]@icloud.com.
    
    An error occurred while evaluating the refund eligibility for your account. Your request has been fully documented and our team will follow up with you shortly to investigate this unauthorized transaction and assist with the refund and cancellation.
    
    Best regards,

And then no one followed up, the conversation was closed without recourse and I wasn't allowed to reply.

I'm not sure how familiar you are with international trading practises, but in multiple jurisdictions, the AI agent assumed legal liability for Anthropic. It accepted that the charge was unauthorized / fraudulent, stated that redressal was needed, but then failed to offer the means to redress it / didn't allow for the refund to continue.

I am not a lawyer, but based on my understanding of prior cases (I read this kind of stuff for fun, don't ask) – in the EU, the US and Canada, users can approach courts and invoke the doctrine of promissory estoppel (again don't quote me on this, IANAL, just like reading case law). And if enough users are affected / do so, it becomes a deceptive practises issue.

I've been thinking about how to solve this problem, and as strange as it sounds, I think Anthropic already has the tools to make the best customer support service in human history. No exaggeration. I think that this crisis could be an opportunity.