Get the research letter
Companies change their terms quietly. We read every version and catch what actually changed. One email a week on the changes that matter and what they mean.
The document discloses that Mistral Medium 3.5 supports agentic and multi-turn conversation workflows via /v1/agents and /v1/conversations endpoints, including built-in tool use accessible through the same routes.
This analysis describes what Mistral AI's agreement states, permits, or reserves. It does not constitute a legal determination about enforceability. Regulatory applicability and practical outcomes may vary by jurisdiction, enforcement context, and individual circumstances. Read our methodology
Agentic endpoints authorize the model to operate in multi-step, tool-using workflows that may involve repeated API calls, external tool invocation, and extended autonomous task execution; the governance implications of agentic deployments, including accountability for model-initiated actions, are operationally distinct from single-turn chat completions.
Interpretive note: The model card does not describe the scope of built-in tools available, the data accessed by those tools, or the oversight mechanisms applicable to agentic task execution, leaving governance implications partially indeterminate from this document alone.
Under this disclosure, the API authorizes deployment of Mistral Medium 3.5 in agentic configurations where the model may execute multi-step tasks, invoke built-in tools, and maintain stateful conversations; downstream deployers are responsible for defining the scope and oversight of agentic task execution within their applications.
Cross-platform context
See how other platforms handle Agentic Workflow and Built-In Tools Endpoints and similar clauses.
Compare across platforms →Monitoring
Mistral AI has changed this document before.
Receive same-day alerts, structured change summaries, and monitoring for up to 20 platforms.
"Agents & Conversations /v1/agents /v1/conversations Built-In Tools /v1/agents /v1/conversationsExcerpt from Mistral AI's Mistral Medium 3.5 Model Card
(1) REGULATORY LANDSCAPE: Agentic AI deployments engage the EU AI Act's provisions on high-risk AI systems and general-purpose AI models where the model takes autonomous actions with real-world consequences; the applicable risk classification depends on the deployment domain. GDPR applies where agentic workflows process personal data, including where the model initiates data retrieval or tool calls. (2) GOVERNANCE EXPOSURE: Medium to High depending on deployment context. Agentic configurations that authorize the model to take actions (tool calls, API invocations, external system interactions) on behalf of users raise accountability and auditability questions that single-turn completions do not. (3) JURISDICTION FLAGS: EU/EEA deployments of agentic AI systems in regulated sectors (healthcare, financial services, legal) face heightened EU AI Act and sector-specific obligations. US federal agency deployments may engage OMB guidance on AI governance for automated decision-making. (4) CONTRACT AND VENDOR IMPLICATIONS: Enterprise contracts deploying agentic workflows should address liability allocation for model-initiated actions, error handling, and human oversight requirements; the model card does not address these operational governance questions. (5) COMPLIANCE CONSIDERATIONS: Organizations deploying Mistral Medium 3.5 in agentic configurations should implement human-in-the-loop oversight mechanisms appropriate to the risk level of the tasks delegated to the model, and should review whether their EU AI Act compliance assessments account for agentic deployment scenarios.
Regulatory citations, enforcement risk, and due diligence action items.
Provision-level monitoring, governance timelines, and regulatory mapping built from archived source documents and historical version tracking.
Agentic endpoints authorize the model to operate in multi-step, tool-using workflows that may involve repeated API calls, external tool invocation, and extended autonomous task execution; the governance implications of agentic deployments, including accountability for model-initiated actions, are operationally distinct from single-turn chat completions.
Under this disclosure, the API authorizes deployment of Mistral Medium 3.5 in agentic configurations where the model may execute multi-step tasks, invoke built-in tools, and maintain stateful conversations; downstream deployers are responsible for defining the scope and oversight of agentic task execution within their applications.
No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by Mistral AI.