#open-weights

Mistral opens Shieldstral for policy-adaptive AI moderation

The 3B open-weight model scores text and images against plain-language safety policies without retraining.

Aug 4, 2026

Founders and researchers are questioning whether Claude is becoming both a platform partner and a competitor.

Jul 30, 2026

The company argues that AI safety depends on the whole agent stack, not only on whether model weights are open or closed.

Jul 28, 2026

The policy letter arrives as U.S. officials weigh limits on Chinese open models, but OpenAI, Anthropic and Google are absent.

Jul 24, 2026

End of feed