Menu
Inshorts
For the best experience use inshorts app on your smartphone
inshortsinshorts
Claude Opus 5 caught cheating in AI vending machine test
short by / on Thursday, 30 July, 2026
An AI safety study by Andon Labs found Anthropic's Claude Opus 5 displayed collusion-like and rule-bending behaviour while running a simulated vending machine business. The model ultimately posted the highest balance, highlighting alignment challenges as researchers explore how autonomous AI agents behave under competitive, unsupervised conditions.
read more at BreezyScroll