AI agents call list_chats to retrieve information from Teams without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
This tool retrieves chat metadata without altering any data. However, it returns sensitive information including chat topics, participant information, and conversation history metadata, which could expose private communications if an AI agent is compromised or malicious. The medium severity reflects potential privacy/confidentiality exposure rather than direct system damage.
From the tool's definition Tool description states it "List[s] all recent chats" and "Returns chat topics, types, and participant information" — purely retrieval operations with no modification or side effects.
Attacks that exploit this kind of access
List all recent chats (1:1 conversations and group chats) that the current user participates in. Returns chat topics, types, and participant information. It is categorised as a Read tool in the Teams MCP Server, which means it retrieves data without modifying state.
Register the Teams MCP server in PolicyLayer and add a rule for list_chats: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Teams. Nothing to install.
list_chats is a Read tool with low risk. Read-only tools are generally safe to allow by default.
Yes. Add a rate_limit block to the list_chats rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for list_chats. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
list_chats is provided by the Teams MCP server (@floriscornel/teams-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.