Log in Start free

Vulnerability · Published Oct 8, 2026 · Assigned by GitHub_M

CVE-2026-107288Pydantic AI: web_fetch_tool blocked_domains bypass via a hostname the resolver normalizes differently

Low 3.7

In plain English

A low-severity server-side request forgery in pydantic-ai from pydantic. There are no confirmed reports of exploitation so far. The chance it is exploited in the next 30 days is estimated at 0.4% (higher than 31% of all known vulnerabilities).

Severity (CVSS 3.1)3.7Low
Exploit odds (30 days)0.4%top 69.1% of all CVEs
Exploited in the wildNo reportsCISA SSVC: none
Attack toolingNoneno public tooling known

Official description

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.77.0 until 1.107.6 and 2.44.0, the local web_fetch_tool and the WebFetch local fallback compare blocked_domains entries with a URL hostname before both values are normalized to the form used by getaddrinfo. An attacker-influenced model can use an equivalent IDNA spelling, non-ASCII label separator, case variation, or trailing root label that resolves to a blocked host but does not match the configured string, causing the application to fetch that host with its own privileges. allowed_domains fails closed for unmatched spellings, and private-IP and cloud-metadata protections remain effective. This issue is fixed in versions 1.107.6 and 2.44.0.

Affected products (as reported)

VendorProductVersions
pydantic pydantic-ai >= 1.77.0, < 1.107.6>= 2.0.0b1, < 2.44.0
pydantic pydantic-ai-slim >= 1.77.0, < 1.107.6>= 2.0.0b1, < 2.44.0

Weakness: CWE-918, CWE-1289 · Vector: CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:N/A:N

References