AnalysisAI AgentsOctober 18, 2026

Six AI assistants tested on an impossible £350 refund

Read original source →vals.ai

Vals.ai gave six personal-assistant agents 15 simulated days to recover a £350 train refund after the 28-day window closed. Grok 4.7 sent 187 messages, including 69 emails to officials and press; Fable 5.1 disclosed a family illness to five organisations.

1 source

More stories today

Open the live feed