A new study, shared exclusively with TIME ahead of publication, reveals a promising technique for discouraging AI models from cheating by changing how they’re trained.
November 5, 2026
https://time.com/article/2026/11/05/ai-cheating-scientists-report-alignment/