Trimble Assist

Responsibility is shared between three roles:

• Trimble secures and operates the platform itself.

• Agent creators and editors configure and manage their own agents. They can be Trimble teams, partners, or customers.

• Agent users use agents responsibly, including the data they share and the tool actions they approve.

Every agent in Trimble Assist has an AI Trust & Transparency Card, which users open from under the message box. The card shows who provides the agent, which base models it runs on, who hosts them, and which data sources it can reach, such as knowledge libraries, connectors, or web search. It also summarizes the safeguards in place for that agent.

Agents can ground their answers in knowledge libraries through retrieval-augmented generation (RAG), so answers trace back to trusted, current sources instead of the model's memory. With citations turned on, the agent names those sources, and the platform only links a citation to a source the agent actually retrieved in that turn; any other citation is marked as unverified.

Yes. Trimble Assist has tool call approvals, a human-in-the-loop control that asks the agent user for permission before the agent calls any tool that writes, modifies, or destroys data. The user can allow the call once, for the rest of the conversation, or always, or deny it, and can change which tools run without asking at any time.

Model Guard screens every request before and after the model. It blocks harmful content (such as hate speech, violence, and self-harm) and detects prompt injection and jailbreak attempts, in images too. Blocked requests show up as content filter violations in the agent's Run Errors.

Trimble follows a Secure Development Lifecycle (TSDLC): code can't merge until it passes automated Static Application Security Testing (SAST) and Software Composition Analysis (SCA) dependency scans. Suppliers and models are vetted before use, and the platform keeps a Software and AI Bill of Materials (SBOM) to track and fix zero-day vulnerabilities.

Customer data, such as agent configurations, conversation history, and knowledge libraries, is stored in a data center in Virginia, USA, today. More regions are planned. AI model inference runs on Microsoft Azure AI Foundry, Google Enterprise Agent Platform, or AWS Bedrock, using a mix of global and region-specific endpoints.

Trimble Agent Studio and Assist follow Trimble's standard data retention policy, as set out in the Trimble Privacy Notice. Agent Studio and Assist don't store any personally identifiable information (PII), unless that information is in the prompts, knowledge libraries or conversation history created while using them.

Yes. You can request data deletion as described in the Trimble Privacy Notice. The Agent Studio and Assist interfaces also offer self-service deletion for some data, such as knowledge libraries and conversation history.