The change

Insightly updated its Terms of Service and Privacy Policy, both now marked “Last Updated: October 14, 2026.” The Terms add a new Section 11B titled “MCP Server and Connected AI Assistants,” effective the same day.

MCP stands for Model Context Protocol, an open protocol for connecting AI assistants to other software. Insightly’s MCP Server lets an assistant such as Claude or ChatGPT work with your CRM data through plain-language prompts, as DestinationCRM reported at launch. The Terms define a Connected AI Assistant as “any third-party artificial intelligence assistant, agent or application that you or a user of your Account connects to the Services through the MCP Server.”

What changed

• Access. “A Connected AI Assistant accesses the Services through the MCP Server using the user credentials, role and permissions of the user of your Account who authorizes it.”

• Responsibility. “Use of the Services by a Connected AI Assistant is deemed use by that user under Section 5.2, and you are responsible for all activity carried out through the MCP Server or by a Connected AI Assistant under your Account.”

• Actions. “A Connected AI Assistant may create and update records in the Services on a user’s behalf.” Whether you are asked to confirm an action first “is controlled by the applicable Connected AI Assistant and its settings, not by Insightly.”

• Output. “Insightly does not guarantee the accuracy, completeness or suitability of any output generated by a Connected AI Assistant, and you bear all risks associated with relying on it.”

• Health data. “The MCP Server is not currently covered by any Business Associate Agreement or Business Associate Addendum between you and Insightly.”

• Suspension. Insightly may limit or end access through the MCP Server “at any time, for any reason or purpose, and in its sole discretion, with or without notice.”

The AI assistant is not Insightly’s subprocessor

When a vendor passes your data to another company to run its service, that company is usually listed as a subprocessor, and the vendor answers for it. Insightly’s Terms put connected assistants outside that model: “Each Connected AI Assistant is a third-party service provided under its provider’s own terms and is not a service provider or subprocessor of Insightly.”

Once the assistant pulls your data, “that data is disclosed at your direction and is then processed under your agreement with that provider.” The Terms add that “Insightly has no visibility into, or control over, how a Connected AI Assistant handles, processes or retains data it receives, and is not responsible for it.” A new paragraph in the Privacy Policy says the same: “we share with that assistant the information it requests on your behalf, within the permissions of your account.”

So the AI provider a user connects becomes a new place your CRM data goes, outside Insightly’s subprocessor list. Checking that provider falls to you: “You are responsible for evaluating whether a Connected AI Assistant meets your legal, privacy and security requirements before connecting it to the Services through the MCP Server.”

That list already names Open AI and Anthropic, the companies behind ChatGPT and Claude, as “subprocessors authorized to process customer data for Insightly services.” When a user connects ChatGPT or Claude through the MCP Server instead, the data goes to the same companies under the user’s own agreement, not Insightly’s.

Outside the SOC 2 report and the BAA

The Privacy Policy used to say “Insightly is SOC 2 compliant and maintains reasonable security measures.” It now says “Insightly maintains a SOC 2 Type II report and reasonable security measures,” followed by a new sentence: “Please note, the Insightly MCP Server was launched after the most recent SOC 2 audit window and is not within the scope of the current report.”

A SOC 2 report is what most security reviews ask a vendor for. If your team approved Insightly on the strength of that report, the MCP Server is a part of the product the report does not cover.

Health data gets a firm line. The Privacy Policy now says: “Please note, the Insightly MCP Server and Insightly’s AI Copilot are not covered by our Business Associate Agreement. Clients subject to HIPAA must not use either of them with Protected Health Information.” A Business Associate Agreement is the contract HIPAA requires before a vendor handles patient data for a healthcare provider or plan. The Terms already kept health data out of AI Copilot, and Section 11B now extends that ban to the MCP Server.

The Privacy Policy also adds a routing note: “Requests made through the Insightly MCP Server may be routed through servers in North America before being directed to the region where your account data is hosted.” For accounts hosted outside North America, that is a new transfer to record.

Turning it off is not one click

The Terms say a user can disconnect an assistant in the assistant’s own settings, but “Disconnecting in this way may not immediately revoke the underlying authorization.” To cut access at once, “you must disable or delete that user in your Account, which will also end that user’s access to the rest of the Services.”

That matters most when something goes wrong, such as an assistant updating records it should not touch. The only immediate stop also locks that user out of the rest of Insightly.

What Insightly commits to

• “Insightly does not use data accessed through the MCP Server to train AI models.”

• MCP usage logs are linked to your user account and cover “the name of each feature requested, the user, email address and account making the request, the time of the request, and whether it succeeded,” and you can ask Insightly to delete them.

• Only a Connected AI Assistant used through the MCP Server is exempt from the Terms’ rules against bots and automated requests. Every other automated route is still banned.

Why this matters

CRM vendors are opening their products to AI assistants, and their terms are placing the risk on customers. HubSpot now makes customers answer for what its AI agents do, and Walmart, Google, and Stripe put AI agent mistakes on you this month. On the other side of the connection, OpenAI’s app terms now cover ChatGPT calling your app on its own. Insightly adds a different point: the assistant is not its subprocessor at all.

For anyone running vendor reviews, an MCP connector is a new kind of data path. Individual users set it up, it sits outside the vendor’s audit and subprocessor list, and it sends data to whichever AI provider the user picks. That is why monitoring the AI vendors in your stack has to cover connectors as well as subprocessor lists.

Potential impact

If your team uses Insightly, the update raises five practical questions:

• Who on your account can connect an AI assistant, and which assistants are they using?

• Do those assistants’ own terms meet your privacy and security requirements, including on training and retention?

• Does your security review rely on Insightly’s SOC 2 report for a feature the report does not cover?

• Is there health data in your CRM that an assistant could reach?

• If an assistant has to be cut off today, who can disable the user, and what else stops when they do?

How Venpo detected it

Venpo monitors Insightly’s legal documents and flagged both updates, with every inserted and deleted line marked. The records are on the Insightly change pages for the Terms of Service and the Privacy Policy. Every quote in this article was checked against the redlines.

Business outcome

Teams that track Insightly’s terms can set rules for AI connectors before users start connecting them, instead of finding out after CRM data has already gone to an outside AI provider. That head start is what continuous monitoring of vendor legal changes is for.

Key takeaway

Since October 14, 2026, an AI assistant connected to Insightly acts as the user who connects it, sits outside Insightly’s subprocessor list, SOC 2 report, and Business Associate Agreement, and may not lose access right away when disconnected, so decide now who on your account may connect one.