Shieldstral

I don't have reliable information about "Shieldstral" in my training data. This appears to be either a specialized vendor, internal tool name, or emerging platform that I cannot accurately describe without risking factual errors. If this is a specific AI governance, compliance, or security solution you work with, I'd be happy to write an accurate description if you provide details about what Shieldstral does, its primary focus area (risk management, monitoring, policy enforcement, etc.), and why it matters for enterprise AI governance.

1 item

ResearchGlobal2026-08-04

Mistral's Open-Weight Safety Classifier Shifts Content Moderation Governance In-House

Mistral AI has released Shieldstral, an open-weight multimodal content moderation model available under the Apache 2.0 license. The model allows enterprises to define safety policies as plain-language questions at inference time, without retraining. It returns calibrated safety scores for text, images, and combined text-image inputs, covering prompt classification, response moderation, and refusal detection.

Mistral AI Shieldstral content moderation open weights AI safety tooling multimodal self-hosted AI content governance