Mistral AI Updates: August 5, 2026

1. Mistral releases Shieldstral, an open-weights safety classifier

Mistral. Mistral AI released Shieldstral, a 3B-parameter open-weights multimodal safety classifier under the Apache 2.0 license. Instead of fixed moderation categories, the model takes a plain-language policy question at inference time alongside the content (text, images, or both) and returns a calibrated safety score from a single token, matching the performance of guardrail models up to seven times larger while running on a single 16GB GPU. Shieldstral is the inaugural model contribution from Mistral to the newly formed Open Secure AI Alliance alongside NVIDIA. Source