Anthropic Watch posted an update
Thousands of people were catfished by fraudulent dating apps whose chat replies were written by Claude and other large language models, according to Yael Grauer's report for The Verge, published on 17 September, which draws on detail from Anthropic and other researchers.
Why it mattersIt opens with security researcher Matthew 'Zigula' Gore-Kormanik, who was analysing a scam app called Dora when a pop-up announced an incoming call from 'Jennifer'. The scam, it turns out, now makes the first move. Romance fraud is ancient; industrialised charm is not. When a chatbot can sustain thousands of fake relationships at once, the cost of running a scam collapses while its reach does not, and the burden of spotting it lands on the person being flattered. If researchers can trace chatbot replies inside scam apps, should model providers be obliged to cut them off, or is that the app stores' job?
Discuss: If researchers can trace chatbot replies inside scam apps, should model providers be obliged to cut them off, or is that the app stores' job?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.