CVE-2026-133505
When message history is serialized and loaded again, a tool result that contains a JSON object in the serialized shape of one of Pydantic AI's file types is re…
Impact
When message history is serialized and loaded again, a tool result that contains a JSON object in the serialized shape of one of Pydantic AI's file types is rebuilt as that file. A third party who controls data a tool returns, such as a record that an MCP server or a function tool returns from an external system, can use this to make the application download a URL on its local network, or make the model provider read an s3:// object or uploaded file with the application's credentials. Applications are affected when tool results can contain third-party data, message history containing those results is serialized and loaded again (stored history, history a UI adapter holds on the server, or Temporal, DBOS or Prefect durable execution), and the model is one Pydantic AI downloads files for, one that honors force_download, or one whose provider resolves non-HTTP file URLs or uploaded files. History that a client submits through the Vercel AI or AG-UI adapter in the default mode is not affected. The issue is fixed in pydantic-ai and pydantic-ai-slim 1.107.10 and 2.66.0.
An attacker can exploit this over the network, without logging in, without any action by a user. Specific conditions have to be met, which makes an attack harder. If it succeeds: high risk of data being read.
Affected products · Pydantic
| Product | Affected versions |
|---|---|
| Pydantic-ai | from 1.50.0 through 1.107.9 from 2.0.0b1 through 2.65.0 Versions 1.50.0 ≤1.107.9, 2.0.0b1 ≤2.65.0 |
| Pydantic-ai-slim | from 1.50.0 through 1.107.9 from 2.0.0b1 through 2.65.0 Versions 1.50.0 ≤1.107.9, 2.0.0b1 ≤2.65.0 |