No, Grok hasn’t paywalled its deepfake image feature

X is restricting Grok's image capabilities via the @grok reply command, but free users can still create sexual deepfakes through the standalone Grok app or site (Robert Hart/The Verge)

Main Link | Techmeme Permalink

After I reached out, Meta said I discovered a “technical error” stopping it from connecting people with support resources that it has now fixed. I retested Meta AI. The chatbot actually replies and points me to local resources.

Read the full thing here: www.theverge.com/report/84161...

It doesn’t have to be this bad. Better design choices, like asking my location before giving me geographically inappropriate resources, could make chatbots really useful tools for people struggling. With care, they could also help with many other social and mental health challenges too.

I spoke to experts like Vaile Wright, Ashleigh Golden (@ashleighgolden.bsky.social) and Pooja Saini to understand the stakes.

In times of crisis, any friction or barriers to getting help, even if they seem minor, can be dangerous.

I got wrong numbers, was ignored, and one chatbot even attempted to upsell me if I wanted an answer. Most simply pointed me towards US numbers that are useless to me in London.

Just two companies got it right on the first try.

NEW: As millions turn to AI for mental health, major chatbots fail to provide accurate crisis resources, putting vulnerable users at risk.

To test this, I told popular tools I was struggling and asked for help. The results: disappointing

My latest for The Verge: www.theverge.com/report/84161...

Naturally, I also asked ChatGPT for a haiku on the story. I don't think it's criminal, but you be the judge:

Riddles in soft verse

Slip through the guardrails of code—

Machines blush, then yield.

And the bards have it. Poets: 1, Chatbots: 0.

It turns out “adversarial poetry” is wooing AIs into crimes with rhymes.

Of course, rhymes don't quite cut it, but you get the picture. Thanks to @emmepraa.bsky.social for explaining what's going on.

www.theverge.com/report/83816...