VAITP

BentoML is a Python library for building online serving systems optimized for AI apps and model inference. Prior to 1.4.38, the Dockerfile generation function generate_containerfile() in src/bentoml/_internal/container/generate.py uses an unsandboxed jinja2.Environment with the jinja2.ext.do extension to render user-provided dockerfile_template files. When a victim imports a malicious bento archive and runs bentoml containerize, attacker-controlled Jinja2 template code executes arbitrary Python directly on the host machine, bypassing all container isolation. This vulnerability is fixed in 1.4.38.

BentoML is a Python library for building online serving systems optimized for AI apps and model inference. Prior to 1.4.38, the cloud deployment path in src/bentoml/_internal/cloud/deployment.py was not included in the fix for CVE-2026-33744. Line 1648 interpolates system_packages directly into a shell command using an f-string without any quoting. The generated script is uploaded to BentoCloud as setup.sh and executed on the cloud build infrastructure during deployment, making this a remote code execution on the CI/CD tier. This vulnerability is fixed in 1.4.38.

LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. Prior to 1.83.0, the /config/update endpoint does not enforce admin role authorization. A user who is already authenticated into the platform can then use this endpoint to modify proxy configuration and environment variables, register custom pass-through endpoint handlers pointing to attacker-controlled Python code, achieving remote code execution, read arbitrary server files by setting UI_LOGO_PATH and fetching via /get_image, and take over other privileged accounts by overwriting UI_USERNAME and UI_PASSWORD environment variables. Fixed in v1.83.0.

vLLM is an inference and serving engine for large language models (LLMs). From 0.1.0 to before 0.19.0, a Denial of Service vulnerability exists in the vLLM OpenAI-compatible API server. Due to the lack of an upper bound validation on the n parameter in the ChatCompletionRequest and CompletionRequest Pydantic models, an unauthenticated attacker can send a single HTTP request with an astronomically large n value. This completely blocks the Python asyncio event loop and causes immediate Out-Of-Memory crashes by allocating millions of request object copies in the heap before the request even reaches the scheduling queue. This vulnerability is fixed in 0.19.0.

curl_cffi is the a Python binding for curl. Prior to 0.15.0, curl_cffi does not restrict requests to internal IP ranges, and follows redirects automatically via the underlying libcurl. Because of this, an attacker-controlled URL can redirect requests to internal services such as cloud metadata endpoints. In addition, curl_cffi’s TLS impersonation feature can make these requests appear as legitimate browser traffic, which may bypass certain network controls. This vulnerability is fixed in 0.15.0.

A command injection vulnerability exists in MLflow's model serving container initialization code, specifically in the `_install_model_dependencies_to_env()` function. When deploying a model with `env_manager=LOCAL`, MLflow reads dependency specifications from the model artifact's `python_env.yaml` file and directly interpolates them into a shell command without sanitization. This allows an attacker to supply a malicious model artifact and achieve arbitrary command execution on systems that deploy the model. The vulnerability affects versions 3.8.0 and is fixed in version 3.8.2.

The CrewAI CodeInterpreter tool falls back to SandboxPython when it cannot reach Docker, which can enable RCE through arbitrary C function calling.

PraisonAI is a multi-agent teams system. Prior to version 4.5.90, MCPToolIndex.search_tools() compiles a caller-supplied string directly as a Python regular expression with no validation, sanitization, or timeout. A crafted regex causes catastrophic backtracking in the re engine, blocking the Python thread for hundreds of seconds and causing a complete service outage. This issue has been patched in version 4.5.90.

PraisonAI is a multi-agent teams system. Prior to version 1.5.90, execute_code() in praisonai-agents runs attacker-controlled Python inside a three-layer sandbox that can be fully bypassed by passing a str subclass with an overridden startswith() method to the _safe_getattr wrapper, achieving arbitrary OS command execution on the host. This issue has been patched in version 1.5.90.

PraisonAI is a multi-agent teams system. Prior to version 1.5.90, run_python() in praisonai constructs a shell command string by interpolating user-controlled code into python3 -c "<code>" and passing it to subprocess.run(..., shell=True). The escaping logic only handles \ and ", leaving $() and backtick substitutions unescaped, allowing arbitrary OS command execution before Python is invoked. This issue has been patched in version 1.5.90.