Access Control Bypass

Overview Affected versions of this package are vulnerable to Access Control Bypass via exposed HTTP endpoints in the cluster network. An attacker can gain unauthorized access to sensitive data and potentially manipulate AI models by sending requests from any co-located pod without authentication...

CVE-2026-59707 LocalAI - Server-Side Request Forgery via POST /models/apply

LocalAI contains an unauthenticated server-side request forgery vulnerability in the POST /models/apply endpoint that allows attackers to fetch arbitrary internal URLs. The endpoint passes unsanitized gallery URL fields directly to gallery.GetGalleryConfigFromURLWithContext without proper...

CVE-2026-59707

LocalAI contains an unauthenticated server-side request forgery vulnerability in the POST /models/apply endpoint that allows attackers to fetch arbitrary internal URLs. The endpoint passes unsanitized gallery URL fields directly to gallery.GetGalleryConfigFromURLWithContext without proper...

CVE-2026-59707 LocalAI - Server-Side Request Forgery via POST /models/apply

LocalAI contains an unauthenticated server-side request forgery vulnerability in the POST /models/apply endpoint that allows attackers to fetch arbitrary internal URLs. The endpoint passes unsanitized gallery URL fields directly to gallery.GetGalleryConfigFromURLWithContext without proper...

CVE-2026-59707 LocalAI - Server-Side Request Forgery via POST /models/apply

LocalAI contains an unauthenticated server-side request forgery vulnerability in the POST /models/apply endpoint that allows attackers to fetch arbitrary internal URLs. The endpoint passes unsanitized gallery URL fields directly to gallery.GetGalleryConfigFromURLWithContext without proper...

CVE-2026-59707

CVE-2026-59707 : LocalAI exposes an unauthenticated SSRF via POST /models/apply. The endpoint forwards unsanitized gallery URL fields to gallery.GetGalleryConfigFromURLWithContext, allowing requests to internal/private/loopback URLs and leaking partial responses in error messages. No exploitation...

PYSEC-2026-2048 Whoogle allows attackers to execute arbitrary code via supplying a crafted search query

An issue in the component /models/config.py of Whoogle search v0.9.0 allows attackers to execute arbitrary code via supplying a crafted search query...

PYSEC-2026-1624 Mesop Class Pollution vulnerability leads to DoS and Jailbreak attacks

From @jackfromeast and @superboy-zjc: We have identified a class pollution vulnerability in Mesop = 0.14.0 application that allows attackers to overwrite global variables and class attributes in certain Mesop modules during runtime. This vulnerability could directly lead to a denial of service Do...

PYSEC-2026-1441 H2O Vulnerable to Arbitrary File Overwrite

In h2oai/h2o-3 version 3.46.0, the /99/Models/name/json endpoint allows for arbitrary file overwrite on the target server. The vulnerability arises from the exportModelDetails function in ModelsHandler.java, where the user-controllable mexport.dir parameter is used to specify the file path for...

PYSEC-2026-1734 Open WebUI stored cross-site scripting (XSS) vulnerability

A stored cross-site scripting XSS vulnerability exists in open-webui/open-webui version 0.3.8. The vulnerability is present in the /api/v1/models/add endpoint, where the model description field is improperly sanitized before being rendered in chat. This allows an attacker to inject malicious...

Open WebUI Allows Arbitrary File Write via the `/models/upload` Endpoint

In open-webui version 0.3.8, the endpoint /models/upload is vulnerable to arbitrary file write due to improper handling of user-supplied filenames. The vulnerability arises from the usage of filepath = f"UPLOADDIR/file.filename" without proper input validation or sanitization. An attacker can...

Open WebUI has SSRF in /openai/models

The /openai/models endpoint in open-webui/open-webui version 0.3.8 is vulnerable to Server-Side Request Forgery SSRF. An attacker can change the OpenAI URL to any URL without checks, causing the endpoint to send a request to the specified URL and return the output. This vulnerability allows the...

PYSEC-2026-1743 Open WebUI has SSRF in /openai/models

The /openai/models endpoint in open-webui/open-webui version 0.3.8 is vulnerable to Server-Side Request Forgery SSRF. An attacker can change the OpenAI URL to any URL without checks, causing the endpoint to send a request to the specified URL and return the output. This vulnerability allows the...

PYSEC-2026-1730 Open WebUI Allows Arbitrary File Write via the `/models/upload` Endpoint

In open-webui version 0.3.8, the endpoint /models/upload is vulnerable to arbitrary file write due to improper handling of user-supplied filenames. The vulnerability arises from the usage of filepath = f"UPLOADDIR/file.filename" without proper input validation or sanitization. An attacker can...

PYSEC-2026-1660 MLFlow unsafe deserialization

Deserialization of untrusted data can occur in versions of the MLflow platform running version 0.5.0 or newer, enabling a maliciously uploaded PyTorch model to run arbitrary code on an end user’s system when interacted with...

PYSEC-2026-1663 MLFlow unsafe deserialization

Deserialization of untrusted data can occur in versions of the MLflow platform running version 1.24.0 or newer, enabling a maliciously uploaded pmdarima model to run arbitrary code on an end user’s system when interacted with...

CVE-2026-55514

CVE-2026-55514 affects the vLLM library (inference/serving) from versions 0.12.0 through older than 0.24.0. Sending a pure prompt embeds payload in a /v1/completions request for a model using M-RoPE triggers an EngineCore assertion, causing a fatal crash that shuts down the entire server applicat...

CVE-2026-55514 vLLM denial of service via prompt embeds on M-RoPE models

vLLM is a library for LLM inference and serving. From 0.12.0 to before 0.24.0, sending a pure prompt embeds payload in a /v1/completions request with a model using M-RoPE causes EngineCore to fail an assertion and fatally crash, shutting down the entire server application. Any remote user who is...

CVE-2026-55646

vLLM is an inference and serving engine for large language models. From 0.22.0 to 0.23.0, the /v1/audio/transcriptions and /v1/audio/translations routes call request.file.read to fully materialize an uploaded audio file into memory before vLLM checks the documented VLLMMAXAUDIOCLIPFILESIZEMB...

PYSEC-2026-972 Core dump when loading TFLite models with quantization in TensorFlow

Impact Certain TFLite models that were created using TFLite model converter would crash when loaded in the TFLite interpreter. The culprit is that during quantization the scale of values could be greater than 1 but code was always assuming sub-unit scaling. Thus, since code was calling...