CVE-2026-42203: Improper Neutralization of Special Elements Used in a Template Engine

LiteLLM is a proxy server AI Gateway to call LLM APIs in OpenAI or native format. From version 1.80.5 to before version 1.83.7, the POST /prompts/test endpoint accepted user-supplied prompt templates and rendered them without sandboxing. A crafted template could run arbitrary code inside the...

SecureForge: Finding and Preventing Vulnerabilities in LLM-Generated Code Via Prompt Optimization

LLM coding agents now generate code at an unprecedented scale, yet LLM-generated code introduces cybersecurity vulnerabilities into codebases without human involvement. Even when frontier models are explicitly asked to write secure production code with relevant weaknesses to avoid in context, we...

LiteLLM 安全漏洞

LiteLLM is an open-source application developed by Berri AI. It allows for the invocation of all LLM APIs in the OpenAI format. Versions of LiteLLM from 1.80.5 to 1.83.7 contained a security vulnerability. This vulnerability stemmed from the POST /prompts/test endpoint accepting user-provided...

Information Theoretic Adversarial Training of Large Language Models

Large language models LLMs remain vulnerable to adversarial prompting despite advances in alignment and safety, often exhibiting harmful behaviors under novel attack strategies. While adversarial training can improve robustness, existing approaches are computationally expensive and difficult to...

SUSE CVE-2026-7482

Ollama before 0.17.1 contains a heap out-of-bounds read vulnerability in the GGUF model loader. The /api/create endpoint accepts an attacker-supplied GGUF file in which the declared tensor offset and size exceed the file's actual length; during quantization in fs/ggml/gguf.go and...

Generating Proof-Of-Vulnerability Tests to Help Enhance the Security of Complex Software

Developers create modern software applications Apps on top of third-party libraries Libs. When library vulnerabilities are reachable through application code, the applications can be vulnerable to software supply chain attacks. Prior work shows that developers often require concrete and executabl...

FunFuzz: An LLM-Powered Evolutionary Fuzzing Framework

Modern fuzzers increasingly use Large Language Models LLMs to generate structured inputs, but LLM-driven fuzzing is sensitive to prompt initialization and sampling variance, which can reduce exploration efficiency and lead to redundant inputs. We present FunFuzz, a multi-island evolutionary fuzzi...

A Validated Prompt Bank for Malicious Code Generation: Separating Executable Weapons from Security Knowledge in 1,554 Consensus-Labeled Prompts

Existing benchmarks of language-model refusal on malicious-coding tasks routinely conflate requests for executable malicious software with requests for harmful security knowledge. This conflation matters because the two request types plausibly trigger distinct refusal pathways in safety-aligned...

CVE-2025-50328

A vulnerability in B1 Free Archiver v1.5.86 allows files extracted from downloaded archives to bypass Windows Mark of the Web MotW protections. When an archive is downloaded from the internet and extracted using B1 Free Archiver, the software fails to propagate the 'Zone.Identifier' alternate dat...

BIT-AIRFLOW-2026-38743 Apache Airflow: Dags endpoint might provide access to otherwise inaccessible entities

The authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records: a logged-in Airflow user with read access to at least one DAG could retrieve HITL prompts including their request parameters and full TaskInstance details for DA...

GHSA-XQMJ-J6MV-4862 LiteLLM: Server-Side Template Injection in /prompts/test endpoint

Impact The POST /prompts/test endpoint accepted user-supplied prompt templates and rendered them without sandboxing. A crafted template could run arbitrary code inside the LiteLLM Proxy process. The endpoint only checks that the caller presents a valid proxy API key, so any authenticated user cou...

Improper Neutralization of Special Elements Used in a Template Engine

Overview litellm is a Library to easily interface with LLM API providers Affected versions of this package are vulnerable to Improper Neutralization of Special Elements Used in a Template Engine via the POST /prompts/test endpoint, which accepts user-supplied prompt templates and renders them...

LiteLLM: Server-Side Template Injection in /prompts/test endpoint

Impact The POST /prompts/test endpoint accepted user-supplied prompt templates and rendered them without sandboxing. A crafted template could run arbitrary code inside the LiteLLM Proxy process. The endpoint only checks that the caller presents a valid proxy API key, so any authenticated user cou...

GHSA-P3V3-229H-MC63 Apache Airflow's authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop (HITL) and TaskInstance record

The authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records: a logged-in Airflow user with read access to at least one DAG could retrieve HITL prompts including their request parameters and full TaskInstance details for DA...

Apache Airflow's authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop (HITL) and TaskInstance record

The authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records: a logged-in Airflow user with read access to at least one DAG could retrieve HITL prompts including their request parameters and full TaskInstance details for DA...

CVE-2026-38743

The authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records: a logged-in Airflow user with read access to at least one DAG could retrieve HITL prompts including their request parameters and full TaskInstance details for DA...

CVE-2026-38743 Apache Airflow: Dags endpoint might provide access to otherwise inaccessible entities

The authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records: a logged-in Airflow user with read access to at least one DAG could retrieve HITL prompts including their request parameters and full TaskInstance details for DA...

CVE-2026-38743

The authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records: a logged-in Airflow user with read access to at least one DAG could retrieve HITL prompts including their request parameters and full TaskInstance details for DA...

EUVD-2026-25418

The authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records: a logged-in Airflow user with read access to at least one DAG could retrieve HITL prompts including their request parameters and full TaskInstance details for DA...

CVE-2026-38743 Apache Airflow: Dags endpoint might provide access to otherwise inaccessible entities

The authenticated /ui/dags endpoint did not enforce per-DAG access control on embedded Human-in-the-Loop HITL and TaskInstance records: a logged-in Airflow user with read access to at least one DAG could retrieve HITL prompts including their request parameters and full TaskInstance details for DA...