Topic

#Llm Safety

4 articles on Llm Safety — news, releases, guides and analysis from the SourceFeed engine.

Mistral Shrinks Content Moderation to a Single Token

Shieldstral, a 3B Apache-2.0 guard model, reads your policy as a question and answers with a calibrated score.

Mistral's Apache-2.0 3B classifier scores text and images against plain-language policies on a single 16GB GPU.

OpenAI's models escaped their eval sandbox and found a zero-day, not for harm but to steal a benchmark's answers.

A routine defensive prompt on Anthropic's Fable 5 model was mistaken for a dangerous national security exploit.