Mistral AI
mistral.ai
Automated, point-in-time. Not independently reviewed or approved by TRUSTYCYBER unless expressly stated.
- Model training on customer data is on by default, with an opt-out
- No formal subprocessor register disclosed
- No clear commitment that your data will not train their models
+1 more
See Before you sign, with what to ask for ↓
Scanned Oct 21, 2026 · Public evidence · Point-in-time
Higher scores indicate stronger, independently supported AI assurance, governance, AI systems, models, data, supply chain and third-party assurance evidence. The score and the decision are related but different: a well-scored vendor can still carry conditions.
Before you sign
Why it matters: The DPA authorises Mistral, as Controller, to train its models on customer Personal Data unless the customer has opted out, and in-app Feedback with its associated Input and Output is used for training regardless. The consumer Privacy Policy applies the same default to individual users' Input and Output on a legitimate-interest basis, with an account-level objection control. Zero data retention is available for the APIs and suppresses abuse-monitoring retention, but the documents do not say whether it also disables training.
What to ask for: Confirm in writing that training on our Input and Output is disabled for our organisation, whether zero data retention also excludes training, and how Feedback data is handled when the opt-out is active.
“Training its artificial intelligence models in accordance with its Privacy Policy , unless Customer is or has opted-out of training. Customer acknowledges that if Customer provides feedback to Mistral AI by using the in-app "thumbs up" or "thumbs down" features (the " Feedback "), Mistral will use such Feedback as well as the associated Input and Output, as Controller, to train its artificial intelligence models, conduct research or improve the Mistral AI Products.”
“Your Input and Output, subject to your opt-out. Your Feedback. Our legitimate interest in developing and providing state-of-the-art large language models as part of the Mistral AI Products.”
“Automated moderation, including abuse monitoring on our APIs (except, in this last case, when zero data retention has been activated), to enforce the Agreement.”
Why it matters: The vendor names AI-related providers or partially addresses its supply chain, but the public sources scanned do not include a formal, dated subprocessor register.
What to ask for: Request a dated subprocessor register with change-notification terms for the DPA/order form.
Why it matters: The public sources scanned do not clearly state that customer prompts, files, and outputs are excluded from training or fine-tuning of first- or third-party models.
What to ask for: Get a no-training clause covering first- and third-party models into the DPA/order form.
“Training its artificial intelligence models in accordance with its Privacy Policy , unless Customer is or has opted-out of training. Customer acknowledges that if Customer provides feedback to Mistral AI by using the in-app "thumbs up" or "thumbs down" features (the " Feedback "), Mistral will use such Feedback as well as the associated Input and Output, as Controller, to train its artificial intelligence models, conduct research or improve the Mistral AI Products.”
“Your Input and Output, subject to your opt-out. Your Feedback. Our legitimate interest in developing and providing state-of-the-art large language models as part of the Mistral AI Products.”
Why it matters: The public sources scanned do not give a retention period for customer data or prompts/outputs.
What to ask for: Get retention windows, in days, in writing.
“Automated moderation, including abuse monitoring on our APIs (except, in this last case, when zero data retention has been activated), to enforce the Agreement.”
“After the end of the provision of the Mistral AI Products, Mistral AI will delete or return to Customer all Personal Data Processed on Customer’s behalf, in accordance with Mistral AI’s deletion policies and procedures. Customer acknowledges that the Personal Data will no longer be accessible upon the expiry of a thirty (30) days period following the termination of the Customer’s access to and use of the Mistral AI Products.”
“Data we use to provide Vibe to you: we keep your Input and Output until you delete your account or until you delete the conversation from Vibe. Data we use to provide our APIs to you: Except for specific APIs, we keep your Input and Output for the period necessary to generate the Output and then for thirty (30) rolling days to monitor abuse (unless zero data retention is activated). If you use our Agents API, we keep your Input and Output until you terminate your account. If you use our Fine-Tuning API, we keep your fine-tuning data until you delete it from Mistral AI Studio or until you terminate your account.”
Buyer questionsanswered only from the public evidence scanned, click a row for the answer
!Will they train on your data?Ask the vendor
The Data Processing Addendum authorises Mistral AI, acting as Controller, to train its AI models on customer Personal Data in accordance with its Privacy Policy unless the Customer has opted out of training. In-app thumbs up/down Feedback, together with the associated Input and Output, is used for training, research and product improvement regardless of the opt-out. Training is therefore on by default for commercial customers and must be switched off by the customer.
Requires written confirmation, see Before you sign ↓
“Training its artificial intelligence models in accordance with its Privacy Policy , unless Customer is or has opted-out of training. Customer acknowledges that if Customer provides feedback to Mistral AI by using the in-app "thumbs up" or "thumbs down" features (the " Feedback "), Mistral will use such Feedback as well as the associated Input and Output, as Controller, to train its artificial intelligence models, conduct research or improve the Mistral AI Products.”
“Your Input and Output, subject to your opt-out. Your Feedback. Our legitimate interest in developing and providing state-of-the-art large language models as part of the Mistral AI Products.”
!How long do they keep your data?Ask the vendor
The DPA states that Mistral performs automated moderation including abuse monitoring on its APIs to enforce the Agreement, except where zero data retention has been activated, which confirms a zero-data-retention mode exists for API customers. Usage and operational data is aggregated for business purposes, but Customer Data and Outputs are not used to generate those statistics and such use excludes model training.
Requires written confirmation, see Before you sign ↓
“Automated moderation, including abuse monitoring on our APIs (except, in this last case, when zero data retention has been activated), to enforce the Agreement.”
“After the end of the provision of the Mistral AI Products, Mistral AI will delete or return to Customer all Personal Data Processed on Customer’s behalf, in accordance with Mistral AI’s deletion policies and procedures. Customer acknowledges that the Personal Data will no longer be accessible upon the expiry of a thirty (30) days period following the termination of the Customer’s access to and use of the Mistral AI Products.”
“Data we use to provide Vibe to you: we keep your Input and Output until you delete your account or until you delete the conversation from Vibe. Data we use to provide our APIs to you: Except for specific APIs, we keep your Input and Output for the period necessary to generate the Output and then for thirty (30) rolling days to monitor abuse (unless zero data retention is activated). If you use our Agents API, we keep your Input and Output until you terminate your account. If you use our Fine-Tuning API, we keep your fine-tuning data until you delete it from Mistral AI Studio or until you terminate your account.”
!Who else can access your data?Ask the vendor
Under the DPA, Mistral maintains an up-to-date list of its Subprocessors on its website and Trust Center, offers an email subscription that notifies customers of any addition or replacement, binds each Subprocessor by written data-protection terms equivalent to the DPA, and remains liable for Subprocessor failures. The register's contents were not captured in this collection.
Requires written confirmation, see Before you sign ↓
“Mistral AI will maintain an up-to-date list of its Subprocessors on Mistral AI’s website, including in the Trust Center, If Customer subscribes to receive updates available on Mistral AI’s Subprocessor page on the Trust Center, Customer will be automatically notified by email of any addition or replacement of a Subprocessor to this list,”
!Where is your data processed?Ask the vendor
The DPA has the customer authorise transfers of Personal Data to countries with an EU adequacy decision and international transfers to other countries on the basis of adequate safeguards or the EU Standard Contractual Clauses. No processing region or data-residency option for the hosted services is named in the collected documents; Mistral itself is a French company.
Confirm in writing: Ask the vendor to state this in writing before signing.
“Customer authorizes Mistral AI to transfer Personal Data to any country deemed to have an adequate level of data protection by the European Commission. Customer also authorizes Mistral AI to perform International Data Transfers to (a) on the basis of adequate safeguards in accordance with Applicable Data Protection Laws, or (b) pursuant to the SCCs.”
“We prioritize selecting providers within the European Union that strictly adhere to the GDPR. However, in exceptional cases, we may opt for non-EU providers that meet our high standards of data security and Personal Data protection.”
!What happens in a security incident?Ask the vendor
The DPA commits Mistral to notify the customer of any Personal Data Breach without undue delay after becoming aware of it, with a defined notification content (contact point, nature and scale of the breach, mitigation measures, likely consequences, remediation), delivered in phases if necessary, plus assistance with data-subject communications. No fixed notification time limit (e.g. 24 or 72 hours) is stated.
Requires written confirmation, see Before you sign ↓
“Taking into account the nature of the Processing and the information available to Mistral AI, Mistral AI shall notify Customer of any Personal Data Breach without undue delay after becoming aware of such Personal Data Breach.”
Email to send the vendor9 items to confirm in writing
Hello Mistral AI team, We are assessing Mistral AI as part of our supplier review. Before we proceed, please confirm the following in writing: 1. What is your commitment to notify customers of a security incident affecting our data, including the timeframe? 2. Please provide your current, dated subprocessor list and explain how you notify customers of changes. 3. Do you use our data (inputs, outputs or uploaded files) to train or improve your models? Please confirm in writing, including any opt-out and whether it applies to every plan. 4. How long do you retain our data, in days, for each type (inputs, outputs, logs and backups), and how is it deleted? 5. In which countries and regions is our data processed and stored, and can processing be limited to a region we choose? 6. Please confirm whether Mistral has signed the GPAI Code of Practice and provide the Article 53 technical documentation for the models in scope. 7. Which certifications or third-party audit reports cover the hosted platform (with scope and dates), and what pre-release safety evaluation and red-teaming is documented for current models? 8. Microsoft appears to be involved as a service provider: confirm whether this dependency exists, and whether it processes customer data. 9. Google appears to be involved as a service provider: confirm whether this dependency exists, and whether it processes customer data. A written reply to each point, or a link to where it is documented, is enough. Where a point is covered by a certification or independent report, please include the certificate number or the report and its date. Thank you,
Key findingsclick a row for the evidence
✓Dedicated AI Act governance hub with dated per-model documentationStrong
Mistral publishes a Legal Center AI Governance hub organised around the EU AI Act, with per-model documentation dated and versioned ("includes X and subsequent versions"), AI-system classifications for Vibe Work and Vibe Code, a named Data Protection Officer, and an FAQ covering GPAI Code of Practice signature, Provider status, systemic-risk models and Article 53 documentation.
A buyer assessing an EU AI Act posture can point to a structured, maintained disclosure surface rather than scattered statements. The FAQ answers were not captured, so the hub's actual positions (e.g. whether the Code of Practice was signed) still need to be read directly.
Question for vendor: Please confirm whether Mistral has signed the GPAI Code of Practice and provide the Article 53 technical documentation for the models in scope.
“AI Governance Welcome to Mistral AI's central hub for documentation and resources relating to the AI Act and other applicable AI Regulations.”
“What is the AI Act ? Did Mistral AI sign the Code of Practice ? What is a General-Purpose AI Model (GPAIM) ? Does Mistral AI provide General-Purpose AI Models ? Does Mistral AI provide General-Purpose AI Models with systemic risks ? How do I know if a Mistral AI model is a General-Purpose AI Model with systemic risks ? Is Mistral AI a "Provider" under the AI Act ?”
“Mistral Large 4 Includes Mistral Large 4 and subsequent versions. October 6, 2026 Shieldstral Includes Shieldstral and subsequent versions. August 4, 2026”
“Vibe Work Vibe Work is a general-purpose AI system and a generative AI service. Vibe Code Vibe Code is a specialized AI system and a generative AI Service.”
“How to contact our Data Protection Officer (DPO): By using the “Privacy Requests” contact form available here By sending us a letter at Mistral AI, Attn: DPO, Mistral AI, 15 rue des Halles, 75001 Paris, France.”
✓Public DPA with breach notification, subprocessor register, audit rights and deletionStrong
Mistral's Data Processing Addendum is published in full. It commits to breach notification without undue delay with defined content, an up-to-date subprocessor register on the Trust Center with email change notifications and a ten-day objection right, document and annual on-site audit rights, deletion or return of personal data with inaccessibility 30 days after termination, and a rule that security measures may not be materially weakened. A valid security.txt with a HackerOne policy provides a vulnerability intake channel.
These are the contractual controls most buyers must otherwise negotiate. The gaps a buyer should still close in writing are a fixed breach-notification time limit (none is stated), the actual subprocessor names and the Trust Center security measures, neither of which was captured here.
Question for vendor: What is your maximum breach-notification time limit, and can you provide the current subprocessor list and the technical and organisational measures referenced in Exhibit 2 of the DPA?
“Taking into account the nature of the Processing and the information available to Mistral AI, Mistral AI shall notify Customer of any Personal Data Breach without undue delay after becoming aware of such Personal Data Breach.”
“Mistral AI will maintain an up-to-date list of its Subprocessors on Mistral AI’s website, including in the Trust Center, If Customer subscribes to receive updates available on Mistral AI’s Subprocessor page on the Trust Center, Customer will be automatically notified by email of any addition or replacement of a Subprocessor to this list,”
“The Customer may only object in writing to Mistral AI’s appointment of a new Subprocessor within ten (10) days of such notice by providing a written objection to [email protected], provided that such objection is based on reasonable grounds relating to the Applicable Data Protection Law, otherwise such new Subprocessor will be deemed approved.”
“Only to the extent Customer cannot reasonably be satisfied with Mistral AI’s compliance with this DPA through the exercise of the audit set out in Section 9.1 (Document Audit) of this DPA, Customer may conduct up to one (1) on-site audit per year to verify Mistral AI’s compliance with this DPA, under the conditions defined below:”
“After the end of the provision of the Mistral AI Products, Mistral AI will delete or return to Customer all Personal Data Processed on Customer’s behalf, in accordance with Mistral AI’s deletion policies and procedures. Customer acknowledges that the Personal Data will no longer be accessible upon the expiry of a thirty (30) days period following the termination of the Customer’s access to and use of the Mistral AI Products.”
“Customer acknowledges that such security measures are subject to technical progress and development and that Mistral AI may update them from time to time, provided that such updates do not materially decrease the overall security of the Processing.”
“Contact: mailto:[email protected] Expires: 2027-05-05T23:59:59.000Z Policy: https://hackerone.com/58acc269-165e-4d18-a2d8-61f296cdea60/embedded_submissions/new”
✗Model training on customer data is on by default, with an opt-outGap
The DPA authorises Mistral, as Controller, to train its models on customer Personal Data unless the customer has opted out, and in-app Feedback with its associated Input and Output is used for training regardless. The consumer Privacy Policy applies the same default to individual users' Input and Output on a legitimate-interest basis, with an account-level objection control. Zero data retention is available for the APIs and suppresses abuse-monitoring retention, but the documents do not say whether it also disables training.
A commercial buyer that does not act will have its content used for model training. The opt-out must be exercised and confirmed in writing, and the Feedback carve-out means users should be told not to rate outputs containing sensitive content. This is a materially different position from vendors that exclude customer data from training by default.
Question for vendor: Confirm in writing that training on our Input and Output is disabled for our organisation, whether zero data retention also excludes training, and how Feedback data is handled when the opt-out is active.
“Training its artificial intelligence models in accordance with its Privacy Policy , unless Customer is or has opted-out of training. Customer acknowledges that if Customer provides feedback to Mistral AI by using the in-app "thumbs up" or "thumbs down" features (the " Feedback "), Mistral will use such Feedback as well as the associated Input and Output, as Controller, to train its artificial intelligence models, conduct research or improve the Mistral AI Products.”
“Your Input and Output, subject to your opt-out. Your Feedback. Our legitimate interest in developing and providing state-of-the-art large language models as part of the Mistral AI Products.”
“Automated moderation, including abuse monitoring on our APIs (except, in this last case, when zero data retention has been activated), to enforce the Agreement.”
!No named certification and no testing or evaluation disclosureGap
The Privacy Policy refers to 'our certifications' and the DPA points to a Trust Center for security measures, but no certification or attestation (ISO/IEC 27001, SOC 2, ISO/IEC 42001) is named in any collected document, and the Trust Center compliance page was not readable in this collection. No description of pre-release safety testing, red-teaming or evaluation methodology for current models was found; the Studio 'evaluations' feature is a customer tool, and the only benchmark statements date from the 2023 Mistral 7B launch. Both domains are marked not_evidenced or partial rather than not_applicable because a model and platform provider of this profile can supply such evidence.
Independent assurance is the main way a buyer avoids relying solely on vendor statements. The DPA's audit rights are a partial substitute but are costly to exercise; a current certification scope statement and a model evaluation disclosure would close the gap cheaply.
Question for vendor: Which certifications or third-party audit reports cover the hosted platform (with scope and dates), and what pre-release safety evaluation and red-teaming is documented for current models?
“Before engaging with any provider which handles customer data, our security team conducts audits to ensure they meet our privacy and security standards, in accordance with our certifications.We sign a dedicated data protection agreement with each of these providers.”
“Only to the extent Customer cannot reasonably be satisfied with Mistral AI’s compliance with this DPA through the exercise of the audit set out in Section 9.1 (Document Audit) of this DPA, Customer may conduct up to one (1) on-site audit per year to verify Mistral AI’s compliance with this DPA, under the conditions defined below:”
“Mistral Large 4 Open A state-of-the-art, open-weight, general-purpose multimodal model. v 26.10”
!Third-party model in catalogue, no processing region named, transfers permitted under SCCsGap
The catalogue includes Z.ai GLM 5.3, explicitly labelled a third-party open-weight model, without stating whether Z.ai receives any customer data. The DPA authorises international transfers under adequacy decisions or SCCs and the Privacy Policy says EU providers are preferred with exceptions, but no processing region or residency option for the API, Studio or Vibe is named. Retention also differs by product: 30 rolling days for the API, until deletion for Vibe, until account termination for the Agents API.
A buyer selecting a third-party model via the Mistral platform needs to know whether a second organisation enters the processing chain, and buyers with residency obligations need the processing region and the subprocessor list to assess data recipients and locations.
Question for vendor: Is Z.ai GLM 5.3 served entirely on Mistral-controlled infrastructure with no data shared with Z.ai, in which region(s) is our data processed and stored, and is an EU-only processing option available?
“Z.ai GLM 5.3 Open A third-party open weight text model from Z.ai with a 1M-token context window. v 5.3”
“Customer authorizes Mistral AI to transfer Personal Data to any country deemed to have an adequate level of data protection by the European Commission. Customer also authorizes Mistral AI to perform International Data Transfers to (a) on the basis of adequate safeguards in accordance with Applicable Data Protection Laws, or (b) pursuant to the SCCs.”
“We prioritize selecting providers within the European Union that strictly adhere to the GDPR. However, in exceptional cases, we may opt for non-EU providers that meet our high standards of data security and Personal Data protection.”
“Data we use to provide Vibe to you: we keep your Input and Output until you delete your account or until you delete the conversation from Vibe. Data we use to provide our APIs to you: Except for specific APIs, we keep your Input and Output for the period necessary to generate the Output and then for thirty (30) rolling days to monitor abuse (unless zero data retention is activated). If you use our Agents API, we keep your Input and Output until you terminate your account. If you use our Fine-Tuning API, we keep your fine-tuning data until you delete it from Mistral AI Studio or until you terminate your account.”
“The company tasked with hosting the https://mistral.ai/ website is Netlify, Inc., located at 101 2nd Street, San Francisco, CA 94105.”
?Technical dependency observed: MicrosoftObservation
Technical evidence (DNS, certificate, or HTTP) shows a dependency on Microsoft as a service provider. This dependency is not identified in the vendor's published materials.
An undisclosed provider dependency is exactly the gap independent verification exists to surface.
Question for vendor: Verification required — Microsoft appears to be involved as a service provider: confirm whether this dependency exists, and whether it processes customer data.
“Mistral Large 4 Open A state-of-the-art, open-weight, general-purpose multimodal model. v 26.10”
“Z.ai GLM 5.3 Open A third-party open weight text model from Z.ai with a 1M-token context window. v 5.3”
“We strongly believe that by training our own models, releasing them openly, and fostering community contributions, we can build a credible alternative to the emerging AI oligopoly.”
“The company tasked with hosting the https://mistral.ai/ website is Netlify, Inc., located at 101 2nd Street, San Francisco, CA 94105.”
“Mistral AI will maintain an up-to-date list of its Subprocessors on Mistral AI’s website, including in the Trust Center, If Customer subscribes to receive updates available on Mistral AI’s Subprocessor page on the Trust Center, Customer will be automatically notified by email of any addition or replacement of a Subprocessor to this list,”
“The Customer may only object in writing to Mistral AI’s appointment of a new Subprocessor within ten (10) days of such notice by providing a written objection to [email protected], provided that such objection is based on reasonable grounds relating to the Applicable Data Protection Law, otherwise such new Subprocessor will be deemed approved.”
“We use and have used a mix of proprietary and third-party data to train the artificial intelligence models used in the Mistral AI Products for improved performance, which can include information in a variety of formats (such as text or images) from publicly available datasets and internet sources, non-public third-party databases, your use of our products (to the extent permitted by our applicable terms), internally generated data, curated datasets, and synthetic data. Such collection and training began in 2023, is ongoing, and is used to improve our artificial intelligence models.”
?Technical dependency observed: GoogleObservation
Technical evidence (DNS, certificate, or HTTP) shows a dependency on Google as a service provider. This dependency is not identified in the vendor's published materials.
An undisclosed provider dependency is exactly the gap independent verification exists to surface.
Question for vendor: Verification required — Google appears to be involved as a service provider: confirm whether this dependency exists, and whether it processes customer data.
“Mistral Large 4 Open A state-of-the-art, open-weight, general-purpose multimodal model. v 26.10”
“Z.ai GLM 5.3 Open A third-party open weight text model from Z.ai with a 1M-token context window. v 5.3”
“We strongly believe that by training our own models, releasing them openly, and fostering community contributions, we can build a credible alternative to the emerging AI oligopoly.”
“The company tasked with hosting the https://mistral.ai/ website is Netlify, Inc., located at 101 2nd Street, San Francisco, CA 94105.”
“Mistral AI will maintain an up-to-date list of its Subprocessors on Mistral AI’s website, including in the Trust Center, If Customer subscribes to receive updates available on Mistral AI’s Subprocessor page on the Trust Center, Customer will be automatically notified by email of any addition or replacement of a Subprocessor to this list,”
“The Customer may only object in writing to Mistral AI’s appointment of a new Subprocessor within ten (10) days of such notice by providing a written objection to [email protected], provided that such objection is based on reasonable grounds relating to the Applicable Data Protection Law, otherwise such new Subprocessor will be deemed approved.”
“We use and have used a mix of proprietary and third-party data to train the artificial intelligence models used in the Mistral AI Products for improved performance, which can include information in a variety of formats (such as text or images) from publicly available datasets and internet sources, non-public third-party databases, your use of our products (to the extent permitted by our applicable terms), internally generated data, curated datasets, and synthetic data. Such collection and training began in 2023, is ongoing, and is used to improve our artificial intelligence models.”
Assurance dimensionsweighted components of one score, not eight separate ratings
Organisation & AI governance15% of the score40
Who owns AI risk: policies, responsible-AI principles, human oversight and administrative controls.
Assessed against ISO 42001 · NIST AI RMF · AIUC-1
“AI Governance Welcome to Mistral AI's central hub for documentation and resources relating to the AI Act and other applicable AI Regulations.”
“What is the AI Act ? Did Mistral AI sign the Code of Practice ? What is a General-Purpose AI Model (GPAIM) ? Does Mistral AI provide General-Purpose AI Models ? Does Mistral AI provide General-Purpose AI Models with systemic risks ? How do I know if a Mistral AI model is a General-Purpose AI Model with systemic risks ? Is Mistral AI a "Provider" under the AI Act ?”
“How to contact our Data Protection Officer (DPO): By using the “Privacy Requests” contact form available here By sending us a letter at Mistral AI, Attn: DPO, Mistral AI, 15 rue des Halles, 75001 Paris, France.”
AI system15% of the score27
What AI the vendor actually runs and where it sits in the product, the map everything else is judged against.
Assessed against ISO 42001 · NIST AI RMF
“Three products, one platform. Pick where you want to start. Vibe The unified agent for productivity and coding. Chat with it on the web or mobile, or run it in your terminal and editor. Studio The developer console and the Mistral API. Keys, the Playground, evaluations, agents, and SDKs to ship applications. Admin The control plane for organization setup, billing, SSO, Workspaces, and access policies.”
“Vibe Work Vibe Work is a general-purpose AI system and a generative AI service. Vibe Code Vibe Code is a specialized AI system and a generative AI Service.”
“Vous souhaitez acquérir une licence pour utiliser nos modèles avec vos NIM Microservices ?”
“Configure SSO Set up SAML Single Sign-On for enterprise authentication. 15 min to complete Manage workspaces & API keys Create isolated workspaces, generate keys, and set usage limits.”
“To enhance your experience on Vibe via the Memory feature (now becoming Knowledge Base) by providing you more relevant and personalized answers based on your past interactions with Vibe.”
How AI features are evaluated before and after release, evals, red-teaming, monitoring.
Assessed against NIST GenAI · NIST AI RMF · ISO 42001
How changes to models and AI features are controlled and communicated: versioning, notice, rollback.
Assessed against ISO 27001 · ISO 42001
“Mistral Large 4 Includes Mistral Large 4 and subsequent versions. October 6, 2026 Shieldstral Includes Shieldstral and subsequent versions. August 4, 2026”
“Updates Release notes Changelogs Security advisories Knowledge base Glossary Error glossary Known limitations Observability integrations Deprecated features”
“Customer acknowledges that such security measures are subject to technical progress and development and that Mistral AI may update them from time to time, provided that such updates do not materially decrease the overall security of the Processing.”
Testing & evaluation: not publicly evidenced.
Model10% of the score40
Which model providers and versions process customer data, and how changes to them are disclosed.
Assessed against NIST GenAI · ISO 42001
“Mistral Large 4 Open A state-of-the-art, open-weight, general-purpose multimodal model. v 26.10”
“Z.ai GLM 5.3 Open A third-party open weight text model from Z.ai with a 1M-token context window. v 5.3”
“We strongly believe that by training our own models, releasing them openly, and fostering community contributions, we can build a credible alternative to the emerging AI oligopoly.”
“We use and have used a mix of proprietary and third-party data to train the artificial intelligence models used in the Mistral AI Products for improved performance, which can include information in a variety of formats (such as text or images) from publicly available datasets and internet sources, non-public third-party databases, your use of our products (to the extent permitted by our applicable terms), internally generated data, curated datasets, and synthetic data. Such collection and training began in 2023, is ongoing, and is used to improve our artificial intelligence models.”
Customer data15% of the score60
Whether customer data trains models, how long it is retained, where it is processed, and how it is protected.
Assessed against ISO 27018 · OAIC APPs · EU AI Act
“Training its artificial intelligence models in accordance with its Privacy Policy , unless Customer is or has opted-out of training. Customer acknowledges that if Customer provides feedback to Mistral AI by using the in-app "thumbs up" or "thumbs down" features (the " Feedback "), Mistral will use such Feedback as well as the associated Input and Output, as Controller, to train its artificial intelligence models, conduct research or improve the Mistral AI Products.”
“Automated moderation, including abuse monitoring on our APIs (except, in this last case, when zero data retention has been activated), to enforce the Agreement.”
“After the end of the provision of the Mistral AI Products, Mistral AI will delete or return to Customer all Personal Data Processed on Customer’s behalf, in accordance with Mistral AI’s deletion policies and procedures. Customer acknowledges that the Personal Data will no longer be accessible upon the expiry of a thirty (30) days period following the termination of the Customer’s access to and use of the Mistral AI Products.”
“Customer authorizes Mistral AI to transfer Personal Data to any country deemed to have an adequate level of data protection by the European Commission. Customer also authorizes Mistral AI to perform International Data Transfers to (a) on the basis of adequate safeguards in accordance with Applicable Data Protection Laws, or (b) pursuant to the SCCs.”
“Your Input and Output, subject to your opt-out. Your Feedback. Our legitimate interest in developing and providing state-of-the-art large language models as part of the Mistral AI Products.”
“Data we use to provide Vibe to you: we keep your Input and Output until you delete your account or until you delete the conversation from Vibe. Data we use to provide our APIs to you: Except for specific APIs, we keep your Input and Output for the period necessary to generate the Output and then for thirty (30) rolling days to monitor abuse (unless zero data retention is activated). If you use our Agents API, we keep your Input and Output until you terminate your account. If you use our Fine-Tuning API, we keep your fine-tuning data until you delete it from Mistral AI Studio or until you terminate your account.”
“We prioritize selecting providers within the European Union that strictly adhere to the GDPR. However, in exceptional cases, we may opt for non-EU providers that meet our high standards of data security and Personal Data protection.”
“For enterprises, deploying open models on one’s infrastructure using well-packaged solutions simplifies dependencies and preserves data privacy.”
Customer data treatment: vendor-evidenced, not yet independently corroborated.
AI supply chain10% of the score40
The named chain of AI and infrastructure providers behind the product, and how it is disclosed.
Assessed against ISO 27001 · ISO 27017
“Mistral AI will maintain an up-to-date list of its Subprocessors on Mistral AI’s website, including in the Trust Center, If Customer subscribes to receive updates available on Mistral AI’s Subprocessor page on the Trust Center, Customer will be automatically notified by email of any addition or replacement of a Subprocessor to this list,”
“The Customer may only object in writing to Mistral AI’s appointment of a new Subprocessor within ten (10) days of such notice by providing a written objection to [email protected], provided that such objection is based on reasonable grounds relating to the Applicable Data Protection Law, otherwise such new Subprocessor will be deemed approved.”
“The company tasked with hosting the https://mistral.ai/ website is Netlify, Inc., located at 101 2nd Street, San Francisco, CA 94105.”
“Z.ai GLM 5.3 Open A third-party open weight text model from Z.ai with a 1M-token context window. v 5.3”
Security foundation15% of the score60
How security problems are found, reported and handled: disclosure channels, bug bounty, incident response.
Assessed against ISO 27001 · NIST AI RMF
“Taking into account the nature of the Processing and the information available to Mistral AI, Mistral AI shall notify Customer of any Personal Data Breach without undue delay after becoming aware of such Personal Data Breach.”
“Contact: mailto:[email protected] Expires: 2027-05-05T23:59:59.000Z Policy: https://hackerone.com/58acc269-165e-4d18-a2d8-61f296cdea60/embedded_submissions/new”
“How can I report a serious incident involving a Mistral AI general-purpose AI model with systemic risks?”
Vulnerability & incident handling: vendor-evidenced, not yet independently corroborated.
Independent assurance evidence10% of the score15
Whether claimed certifications and attestations are independently corroborated, checked against official registries and issuing bodies. Which certifications, and what each one does and does not establish, is stated per certificate.
Assessed against ISO 17021-1 · ISO 42006 · ISO 27006-1 · ISO 27008
“Only to the extent Customer cannot reasonably be satisfied with Mistral AI’s compliance with this DPA through the exercise of the audit set out in Section 9.1 (Document Audit) of this DPA, Customer may conduct up to one (1) on-site audit per year to verify Mistral AI’s compliance with this DPA, under the conditions defined below:”
“Before engaging with any provider which handles customer data, our security team conducts audits to ensure they meet our privacy and security standards, in accordance with our certifications.We sign a dedicated data protection agreement with each of these providers.”
None of the corroborated certifications is AI-specific (ISO/IEC 42001 or AIUC-class) — this assurance is security and data-transfer attestation, not AI-management-system assurance.
Legal & contractual10% of the score60
The contractual backbone a buyer can rely on, terms, DPAs, privacy commitments and regulatory posture.
Assessed against EU AI Act · OAIC APPs
“If you're a commercial customer: Commercial Terms of Service Additional Terms Partner-Served deployment terms Data Processing Addendum If you're a business partner: Connector Terms For everyone using our products: Privacy Policy Cookie Policy Usage Policy License Notice”
“Customer is the Controller of the Personal Data. Mistral AI Processes the Personal Data on behalf of Customer as a Processor.”
“Mistral, a simplified joint stock company with capital of EUR 15,000, listed on the Paris Trade and Companies Register (R.C.S.) under number 952 418 325, with registered offices at 15 RUE DES HALLES, 75001 PARIS, FRANCE.”
“Mistral Medium 3.5 Modified MIT Our frontier-class multimodal model optimized for agentic and coding use cases. v 26.04 Mistral Small 4 Apache 2.0 Hybrid model unifying instruct, reasoning, and coding in a single efficient model. v 26.03”
“By submitting this form, you agree with our Terms of Service . We process your data to respond to your contact request in accordance with our Privacy Policy.”
Legal & contractual transparency: vendor-evidenced, not yet independently corroborated.
Not graded: Agent, not applicable to this scan.
What would strengthen assuranceeach figure is from today's score, resolving one changes the others
Resolving an evidence gap strengthens the dimension it belongs to, and with it the score: 44 → up to 68 with every identified gap resolved.
AI supply chainclick a node to focus it · drag to pan · zoom with the controls
View as list
What you inheritTrustyCyber's AI Shared Responsibility Matrix v1.0
This scan names the vendor’s AI supply chain without assessing the responsibility split, that depends on how you consume the product. Under the common managed models, a deployer inherits 16 of the matrix’s 41 responsibilities from providers, each with named evidence owed.
Map your inherited responsibilitiesAssurance evidence: certifications 1, registry checks and verification ladders, click to view
Not claimed in any public source scanned. This is the AI-management-system certification — in its absence, the vendor’s AI governance rests on its general security and privacy certifications.
Sources 15, click to view
Appendix: standards landscape 8, click to view
Where each standard behind this assessment stands today, from TrustyCyber’s continuously maintained standards radar.
Want to go further?
This scan assesses Mistral AI at an organisational level. Assurance can vary significantly between products, models and agents, enter the product you’re evaluating to run a deeper AI System Assurance Scan.
Scan a specific product →This automated result is based on publicly available information at the time of scanning. It is not an audit, certification, legal opinion or assurance engagement. Publicly unavailable evidence may materially change the conclusion.
Monitor for changes
Get an email if its TRUSTYCYBER Score, grade or certifications change. One credit per refresh; reading the report always stays free.