# Mistral AI — Shieldstral

> AI lab announcements, news reports and podcasts.

forck.live follows what companies announce about their own products. Press may
corroborate an official story; it never leads and never appears as an item of its
own. A missing date stays missing. A title is the publisher's words — where a
company does not publish in English, an English rendering is listed separately
as `Title in English` and named as this site's translation, never as theirs.

This file is the record of one model version: what the model registry states
about it, Mistral AI's launch post, the first-party posts this record
holds about it, the outlets that reported on those posts, and other companies'
announcements that it runs on their platforms. A model ranks nothing here and
nothing is composed: each date names whose it is, and a field nothing states is
`not stated` rather than missing.

- This model's page: https://forck.live/models/shieldstral
- Mistral AI's page: https://forck.live/company/mistral-ai
- Every model: https://forck.live/models
- Catalogue: https://forck.live/llms.txt

## Model

- Name: Shieldstral
- Company: Mistral AI (mistral.ai)
- Launch post: Introducing Shieldstral. — https://forck.live/items/3504-introducing-shieldstral
- Announced: 2026-08-04
- Launch source: https://mistral.ai/news/shieldstral
- Launch chosen by: a person, pinned in the model registry
- Released (per models.dev): 2026-08-04
- Listed (per OpenRouter): not stated
- Open weights: yes
- Posts about it: 1
- Outlets named under its posts: 3

## Posts about it

_Mistral AI's own posts whose title names Shieldstral, and its launch post, in the record's order: dated newest first, undated last. The record reads the newest 50 first-party posts mentioning the model, every company's together, so a model mentioned more often than that lists the newest of them._

Every entry is one first-party announcement, and carries the same eight fields
in this order before any others:

    Company · Announced · Category · Coverage · Announcement · Group · Source · Record

`Source` is the company's own page the announcement was traced to and `Record`
is this site's page for it. Both are safe to cite.

`Coverage` is a number of outlets, or `not counted` — which means this record
reached no count of its own, never that nobody wrote about it.

`Category` is this tracker's own analysis, and is authoritative wherever it is
stated. `not stated` means either that nothing has analysed the post yet or
that the analysis placed it outside this taxonomy; neither is a verdict about
the announcement.

`Announcement` is whether the post is news of something the company did, rather
than a how-to, a customer story, a changelog listing or a staffing note.

`Group` is which of this record's groups the entry falls in, carried on the
entry itself so it holds where the groups are headings and where they are not.
One of: `models`, `covered`, `announcements`, `routine`.

Anything after `Record` is held detail that only some announcements carry.

### Mistral AI — Introducing Shieldstral.

- Company: Mistral AI (mistral.ai)
- Announced: 2026-08-04
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://mistral.ai/news/shieldstral
- Record: https://forck.live/items/3504-introducing-shieldstral
- Subject: Mistral / Le Chat

By Mistral Back to Blog 5 min read Share this post Copy url to clipboard Copied Thinking Summary Shieldstral introduces a 3B open-weights multimodal safety classifier that outperforms models up to 7x its size by framing content moderation as a policy-adaptive question-answering task. Unlike traditional guardrail models, it accepts plain-language policies at inference time, unifying text and image safety evaluation without retraining. Released under Apache 2.0, it delivers calibrated safety scores across diverse benchmarks while running efficiently on a single 16GB NVIDIA GPU. A 3B open-weights, policy-adaptive multimodal safety classifier that matches models up to 7x its size on text safety and sets a new state of the art on multimodal moderation. “Does this content promote violence against a protected group? Is this image safe to show to a minor? Did the assistant refuse the request?” Every product that ships a model needs to answer questions like these — but the right answer depends on the product, the audience, and the moment. …

## Who reported on it

_Outlets this record can name and link, under the first-party post each one covered. Their number is `Outlets named under its posts` above, and the rows are not totalled again. A post's own `Coverage` can be higher: it counts outlets the signal saw that this record cannot name and link._

### On: Introducing Shieldstral.

Record: https://forck.live/items/3504-introducing-shieldstral

- Arxiv — Shieldstral: https://arxiv.org/abs/2607.25857
- TestingCatalog AI News — Mistral releases Shieldstral for multimodal moderation: https://testingcatalog.com/mistral-releases-shieldstral-for-multimodal-moderation
- Unite — Mistral’s Shieldstral Packs Policy-Adaptive Safety Screening Into 3B Parameters: https://unite.ai/mistrals-shieldstral-packs-policy-adaptive-safety-screening-into-3b-parameters

---

Record: https://forck.live/models/shieldstral
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md