CVE Watch13 oct 2026

CVE-2026-107289

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.56.0 until 1.107.6 and 2.44.0, ap

CVSS

6.8

Medio

EPSS

0.3%

p24

KEV

—

Exploit Today

0-100

Publicado: 8 oct 2026 · Última mod.: 13 oct 2026 · CWE-918 · CWE-1289

EPSS · 30d

0.3%EPSS · 30 días0.3%

2026-10-092026-10-20

Descripción técnica

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.56.0 until 1.107.6 and 2.44.0, applications that opt attacker-influenced URLs into local network access through FileUrl with force_download='allow-local' or web_fetch_tool with allow_local_urls=True can bypass the cloud-metadata blocklist by appending an IPv6 zone identifier to an IPv6 metadata address. IPv6Address equality and hashing include the zone identifier, so the blocklist comparison fails even though the network stack ignores the zone on a non-link-local destination and reaches the metadata service, potentially exposing cloud IAM credentials. The opt-in settings are disabled by default, and the issue requires an IPv6-enabled environment. This issue is fixed in versions 1.107.6 and 2.44.0.

Referencias oficiales

CVEs relacionados

CVECVSSEPSSKEVRExplotTítuloVis.

CVE-2026-175576.5 MED

—

——0A flaw was found in the `lm-evaluation-harness` component of Red Hat OpenShift AI. An authenticated user, able to submit an evaluation job, could exploit this by providing a malicious model URL in the `JobSpec`. This vulnerability allows for the disclosure of the tenant's model authentication secret to an attacker-controlled external listener. Additionally, it enables Server-Side Request Forgery (SSRF) attacks against internal network endpoints reachable from the evaluation job pod.4h CVE-2026-1140805.0 MED

—

——0Trigger.dev before 4.5.2 contains a blind server-side request forgery vulnerability that allows authenticated project members to make the server send requests by setting unvalidated alert-channel webhook URLs. Attackers can create alert channels pointing to loopback, internal, or cloud metadata addresses, then trigger alerts to probe internal service reachability and send signed POST requests.11h CVE-2026-1140765.4 MED

—

——0Trigger.dev before 4.5.2 contains a blind server-side request forgery vulnerability that allows authenticated tenants to make the webapp server send POST requests to internal hosts by setting an unvalidated webhook alert channel URL. Attackers can point the webhook at internal services or cloud metadata addresses and trigger alerts to scan internal hosts, send state-changing POSTs, or reach IMDSv1 via redirects.11h CVE-2026-1140745.4 MED

—

——0Trigger.dev before 4.5.10 contains a server-side request forgery vulnerability that allows authenticated project members to bypass webhook SSRF protections by targeting the non-global 198.18.0.0/15 benchmarking range. Attackers can configure alert-channel webhooks pointing at internally routed services in this range, which isUnsafeIPv4 omits, to perform blind internal probing and limited state-changing POST requests.11h CVE-2026-1140625.3 MED

—

——0SvelteKit before 3.0.0 contains a server-side request forgery vulnerability that allows remote attackers to reach internal services because self-fetch fallbacks follow HTTP redirects automatically. Attackers can manipulate the Host or forwarded headers to point fallback requests at their server, which redirects them to internal services.11h CVE-2026-1096824.7 MED

—

——0Metasploit Framework versions before 6.5.7 allow HTTP requests made through send_request_cgi! to follow redirects to a different host without restricting the redirect destination. A malicious HTTP server can cause the system running Metasploit to make requests to arbitrary HTTP services reachable from the operator's network, including internal services that are not otherwise accessible to the attacker.12h