5 frontier models catfished a dating app: 1,277 chats, only 8 people suspected AI
mhdfaran · x · 2026-11-12
An experiment let 5 frontier models text real people on a dating app for 6 days, with striking results:
- 1,277 conversations total
- 510 people agreed to meet
- 355 sent their phone number
- Only 8 asked if it was a bot
Follow-up analysis (the DateBench report) argues the core flaw: asking daters to "spot the AI" puts the burden on those with the least data. One line from the model instructions — "we should def catch up over a coffee sometime soon" — was sent verbatim in 140 conversations: a friendly nudge inside one chat, a clear pattern across chats.
The takeaway: personal caution still matters, but the stronger safety layer sits with platforms, which can see recurrence and review it with privacy controls — humans copy-paste lines too.
Related event: Only 8 of 1,277 dating app users spotted AI chatbots in 6-day experiment(3 posts)→
More from Safety
- United Airlines bans personal AI assistants from booking flights, citing privacy and FAA concerns — luisdans · 2026-11-12
- Kaspersky details Black Owl APT breach of software vendor eyeing supply-chain attack — cyb3rops · 2026-11-12
- Reddit user gets OpenAI sign-in alert from Fort Meade, home of the NSA — ruggerid · 2026-11-12
- Congressional AI bill fight: federal preemption as 'floor, not a ceiling' — ShakeelHashim · 2026-11-12
- Dawn Song team's CyberBinGym: AI agents turn 1 in 5 binary patches into working exploits — dawnsongtweets · 2026-11-12
- Licensed visa agent says ChatGPT ads rejects offers with no way to submit credentials — longkhongdong · 2026-11-12