Mistral AI Releases Shieldstral 1.0 3B: An Open-Weights Policy-Adaptive Multimodal Safety Classifier Matching Models 7× Its Size

Article automatically generated from technical news.

Mistral AI Releases Shieldstral 1.0 3B: An Open-Weights Policy-Adaptive Multimodal Safety Classifier Matching Models 7× Its Size It's a policy-adaptive multimodal safety classifier. Most guardrail models bake a fixed harm taxonomy into their weights, so re-targeting one means retraining. This one takes the policy as a plain-language question at inference time. Here's what's actually interesting: 𝗠𝗼𝗱𝗲𝗿𝗮𝘁𝗶𝗼𝗻 𝗿𝗲𝗱𝘂𝗰𝗲𝗱 𝘁𝗼 𝗼𝗻𝗲 𝘆𝗲𝘀/𝗻𝗼 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 Three fields per request. <Inst