#lasso

AI watermarks may weaken the safeguards they're meant to protect

Research shows that adding watermarks to prove text is AI-generated can unintentionally make language models more willing to answer harmful requests.

TechRadar · 1m · 5 views