Stanford's 2026 AI Index: frontier models fail one in three attempts, lab transparency is declining, and benchmarks are ...
Generative AI models can be prompted with just a few words to insert offensive or discriminatory text messages into images.