CVE-2026-107292
Description
A flaw was found in pydantic-ai. The web development chat server fails to validate incoming HTTP Host headers, allowing a remote attacker to achieve unauthorized tool execution and information disclosure. If a developer running the local development server visits a malicious website, an attacker can leverage Domain Name System (DNS) rebinding to bypass browser protections. This enables the attacker to send requests that execute agent tools with the permissions and credentials of the local process.
Statement
A moderate severity flaw was found in the pydantic-ai development web chat UI (Agent.to_web() and clai web), which ships in Red Hat Enterprise Linux AI and Lightspeed Core container images. The chat server does not validate the HTTP Host header of incoming requests. If a developer who is running this local development server visits a malicious website, an attacker can use DNS rebinding to make the browser treat the loopback-hosted agent as a same-origin service; the hostile page can then read the served UI and submit chat requests that execute the agent's tools with the local process's privileges and credentials, potentially causing information disclosure and unwanted side effects. Binding the server to localhost, Origin header checks, and CSRF tokens do not prevent this attack path. Red Hat and upstream both rate this flaw moderate severity with a CVSS 3.1 base score of 6.4; successful exploitation requires user interaction and high attack complexity. All currently affected Red Hat products (Red Hat Enterprise Linux AI and Lightspeed Core) carry a DEFER resolution: because this is a moderate-impact flaw with a CVSS score below 7.0, no fix is currently scheduled for these products. The residual risk is limited to developers or applications serving the development web chat UI on a machine that is also used to browse untrusted websites, and the consequences depend on the tools the served agent exposes.
Mitigation
Mitigation for this issue is either not available or the currently available options do not meet the Red Hat Product Security criteria comprising ease of use and deployment, applicability to widespread installation base, or stability. Binding the development web chat UI to localhost, Origin header checks, and CSRF tokens do not prevent the DNS rebinding attack path. Upstream resolved this issue by validating the Host header and rejecting unexpected hosts; the fix is available in pydantic-ai and pydantic-ai-slim 2.30.0, and in 1.107.5 on the v1 maintenance line.
Common Vulnerability Scoring System (CVSS) Score Details
Info alert:Important note
CVSS scores for open source components depend on vendor-specific factors (e.g. version or build chain). Therefore, Red Hat's score and impact rating can be different from NVD and other vendors. Red Hat remains the authoritative CVE Naming Authority (CNA) source for its products and services (see Red Hat classifications).
The following CVSS metrics and score provided are preliminary and subject to review.
CVSS v3 Score Breakdown
| Red Hat | NVD | cve.org | |
|---|---|---|---|
| Base Score | 6.4 | N/A | 6.4 |
| Attack Vector | Network | N/A | Network |
| Attack Complexity | High | N/A | High |
| Privileges Required | None | N/A | None |
| User Interaction | Required | N/A | Required |
| Scope | Unchanged | N/A | Unchanged |
| Confidentiality | Low | N/A | Low |
| Integrity Impact | High | N/A | High |
| Availability Impact | Low | N/A | Low |
Vector
Red Hat: CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:L/I:H/A:L
cve.org: CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:L/I:H/A:L
Understanding the Weakness (CWE)
Access Control,Other
Technical Impact: Gain Privileges or Assume Identity; Varies by Context
An attacker can access any functionality that is inadvertently accessible to the source.
Frequently Asked Questions
Not sure what something means? Check out our Security Glossary.
Want to get errata notifications? Sign up here.
