Gmail AI hallucinates, distorts email contents | Hacker News Reader

Gmail AI hallucinates, distorts email contents

t-online.de

I smell class action. The first time smart mail hallucinates something like an appointment cancellation or flight changes which leads to a real world loss, across a big enough surface of users, lawyers are going to be circling.

If this causes corporate loss, it's even more likely. GSuite is behind a lot of things.

I bet they have disclaimer in the ToS

Wouldn't that turn out to be a candle in a hurricane situation given the level of risk involved?

I didn't find one about generative AI, but I suppose this is it:

    To the extent allowed by applicable law:
    - Google is liable only for its breaches of these terms or applicable service-specific additional terms
    - Google isn’t liable for:
      - loss of profits, revenues, business opportunities, goodwill, or anticipated savings
      - indirect or consequential losses
      - punitive damages

https://policies.google.com/terms (this is titled "Google Terms of Service" & is presumably broader but is what I got by going to Gmail and clicking "Terms".)

The terms may be different for corporate users.

Disclaimers are only as strong as the bank balance behind or defending the assault. Loss of custom typically follows a breakdown in trust as well. I suspect anyone seeking to fall back on ToS would find reasons to discuss matters round a table rather than test the strength at the bar.

I'm not a lawyer.

> Disclaimers are only as strong as the bank balance behind or defending the assault.

Googles bank balance is, ummmm, pretty good I think.

I don't expect these hallucinations would impact enough people to cause a material loss of custom,such that google would even notice, let alone care. If you're the person affected, you'll be on your own.

If it was widespread enough to cause loss of custom, they'd 'simply' fix the issue before it got to that point.

So does "AI overviews".

I guarrantee you that you're now having to waste more time double checking even the sources hallucinated from these AI systems.

Reply on news.ycombinator.com