Hugging Face Datasets versions up to 5.1.0 contain a deserialization vulnerability in the deprecated pandas packaged builder. This flaw allows attackers to execute arbitrary Python code by supplying a malicious pickle file that is loaded without validation. When a victim loads such a dataset using datasets.load_dataset("pandas", data_files=...), the embedded payload executes with the victim process's privileges.
tokenizers (Hugging Face) is affected by an out-of-bounds buffer access in BpeBuilder::build (tokenizers/src/models/bpe/model.rs). When loading a tokenizer.json via Tokenizer::from_file/from_str, the builder sizes a scratch buffer to the longest vocabulary key, then writes each concatenated merge rule into it. A merge whose concatenated token exceeds the longest vocabulary key overruns the buffer, which Rust turns into a panic that aborts the process in Rust and FFI embeddings. This occurs at load time with no encoding required, so an attacker who supplies a crafted tokenizer.json can cause a denial of service. A secondary defect at the same location can cause a usize underflow (panic in debug, potential memory corruption in release) when continuing_subword_prefix is set and a merge token is shorter than the prefix. Observed in version 0.23.1.
CVE-2026-75104 is a path traversal vulnerability in the Hugging Face Transformers library. It arises because shard filenames in checkpoint index files are not properly validated, allowing crafted index files with directory traversal sequences or absolute paths. This can enable unauthorized reading of arbitrary files outside the intended model directory. The vulnerability affects all versions of the transformers library and has a CVSS score of 6.8, indicating medium severity.
Hugging Face Accelerate through 1.14.0 contains a path traversal vulnerability in load_checkpoint_in_model and load_checkpoint_and_dispatch functions that fail to sanitize weight_map entries from sharded checkpoint indexes. Attackers can supply relative paths with ../ sequences or absolute paths to read arbitrary files, or point shard entries at named pipes to cause indefinite blocking and denial of service.
Hugging Face peft's LoRA-GA and CorDA initialization modules (src/peft/tuners/lora/corda.py lines ~102 and ~163, and src/peft/tuners/lora/loraga.py line ~101) call torch.load on config-specified cache/covariance files without weights_only=True, bypassing peft's own safe-loading wrapper used elsewhere in the codebase.
A vulnerability in huggingface/transformers versions < 5.10.0 allows an attacker to perform arbitrary file writes via path traversal. The issue resides in the `save_pretrained()` methods of `PreTrainedTokenizerBase` and `ProcessorMixin`, where keys from the `chat_template` dictionary are used directly as filenames without proper validation. An attacker can exploit this by publishing a malicious Hugging Face Hub repository with a crafted `tokenizer_config.json` file. When a victim downloads and saves the tokenizer or processor, the attacker-controlled keys can escape the intended save directory, enabling arbitrary file writes with attacker-controlled content. This vulnerability affects multiple processors inheriting from `ProcessorMixin`, including Idefics, Florence, Gemma, Phi, and Qwen-VL.
Datasets through 5.0.0, fixed in commit f989ef9, contains a path traversal vulnerability in folder-based dataset builders where the file_name metadata field is not properly validated before being joined to the dataset directory. Attackers can supply crafted file_name values with directory traversal sequences to read arbitrary local files, which are then embedded into output when save_to_disk or push_to_hub is called.
Datasets through 5.00, fixed in commit ad2d853, contains a symlink-following vulnerability in Extractor.extract() that allows local attackers to write arbitrary files by pre-planting symlinks at predictable output paths. Attackers can redirect archive extraction to arbitrary filesystem locations in shared-cache environments, enabling overwrite of sensitive files and potential privilege escalation or code execution.
Diffusers through 0.39.0, fixed in commit cee298c, contains a path traversal vulnerability in the _get_checkpoint_shard_files function that allows attackers to read arbitrary files by supplying malicious weight_map values in model index JSON. Attackers can use ../ sequences or absolute paths in weight_map entries to escape the model directory and read safetensors files outside the intended location during model loading.
text-generation-inference through 3.3.7 contains a server-side request forgery (SSRF) vulnerability in the OpenAI-compatible multimodal chat completions endpoint that allows unauthenticated network attackers to coerce the server into issuing arbitrary HTTP GET requests by supplying a crafted image_url value in chat message content. The fetch_image function in router/src/validation.rs performs no validation of private, loopback, link-local, or cloud metadata target addresses, and the reqwest HTTP client follows redirects by default, enabling attackers to bypass scheme checks via redirect chains to reach internal services and cloud instance-metadata endpoints for internal port scanning and credential theft.
Diffusers is the a library for pretrained diffusion models. Prior to 0.38.0, Diffusers' DiffusionPipeline.from_pretrained flow can bypass the trust_remote_code guard because download() validates model_index.json and custom pipeline code before later loading from a cached folder that can change, allowing a Hub repository with custom .py pipeline code to execute through the custom pipeline flow without passing custom_pipeline or trust_remote_code=True. This issue is fixed in version 0.38.0.
A vulnerability in the LightGlue model loading path of huggingface/transformers version 5.2.0 allows an attacker-controlled model repository to execute arbitrary code during model initialization. The issue arises because the `trust_remote_code` parameter, intended to prevent remote code execution, is overridden by untrusted serialized configuration data in a nested code path. Specifically, when loading a LightGlue model using `AutoModel.from_pretrained()` with `trust_remote_code=False`, the `LightGlueConfig` reads the `trust_remote_code` value from the untrusted `config.json` file and propagates it into nested `AutoConfig.from_pretrained()` calls. This results in the execution of attacker-provided Python modules, even when the victim explicitly disables remote code execution. The vulnerability poses a high risk for environments such as API inference servers, research notebooks, CI/CD pipelines, and model evaluation workers, potentially leading to credential theft, lateral movement, or persistence/backdoor deployment.
A critical remote code execution vulnerability exists in all versions of the HuggingFace transformers library prior to version 5.3.0. The vulnerability allows an attacker to craft a malicious `config.json` file containing the `_attn_implementation_internal` field set to an attacker-controlled HuggingFace Hub repository ID. When a victim loads this model using the standard `AutoModelForCausalLM.from_pretrained()` API, the library downloads and executes arbitrary Python code from the attacker's repository with the victim's full OS privileges. This issue arises due to unfiltered deserialization of configuration attributes, insufficient sanitization of internal fields, and unsandboxed execution of downloaded kernels. The vulnerability bypasses the `trust_remote_code` security mechanism, is invisible to the victim, and exploits the standard documented usage pattern, making it particularly severe. Users are advised to upgrade to version 5.3.0 or later to mitigate this issue.
Diffusers is the a library for pretrained diffusion models. Prior to 0.38.0, diffusers 0.37.0 allows remote code execution without the trust_remote_code=True safeguard when loading pipelines from Hugging Face Hub repositories. The _resolve_custom_pipeline_and_cls function in pipeline_loading_utils.py performs string interpolation on the custom_pipeline parameter using f"{custom_pipeline}.py". When custom_pipeline is not supplied by the user, it defaults to None, which Python interpolates as the literal string "None.py". If an attacker publishes a Hub repository containing a file named None.py with a class that subclasses DiffusionPipeline, the file is automatically downloaded and executed during a standard DiffusionPipeline.from_pretrained() call with no additional keyword arguments. The trust_remote_code check in DiffusionPipeline.download() is bypassed because it evaluates custom_pipeline is not None as False (since the kwarg was never supplied), while the downstream code path that actually loads the module resolves the None value into a valid filename. An attacker can achieve silent arbitrary code execution by publishing a malicious model repository with a None.py file and a standard-looking model_index.json that references a legitimate pipeline class name, requiring only that a victim calls from_pretrained on the repository. This vulnerability is fixed in 0.38.0.
Diffusers is the a library for pretrained diffusion models. Prior to 0.38.0, a trust_remote_code bypass in DiffusionPipeline.from_pretrained allows arbitrary remote code execution despite the user passing trust_remote_code=False (or omitting it, which is the default). The vulnerability has three variants, all sharing the same root cause — the trust_remote_code gate was implemented inside DiffusionPipeline.download() rather than at the actual dynamic-module load site, so any code path that bypassed or short-circuited download() also bypassed the security check. DiffusionPipeline.from_pretrained('repoA', custom_pipeline='attacker/repoB', trust_remote_code=False) — the gate evaluated against repoA's file list rather than repoB's, so repoB's pipeline.py was loaded and executed. DiffusionPipeline.from_pretrained('/local/snapshot', custom_pipeline='attacker/repoB', trust_remote_code=False) — the local-path branch never invoked download(), so the gate was never reached and remote code from repoB executed. DiffusionPipeline.from_pretrained('/local/snapshot', trust_remote_code=False) where the snapshot contains custom component files (e.g. unet/my_unet_model.py) referenced from model_index.json — same root cause; the local path skipped download() and custom component code executed. This vulnerability is fixed in 0.38.0.
A vulnerability in the HuggingFace Transformers library, specifically in the `Trainer` class, allows for arbitrary code execution. The `_load_rng_state()` method in `src/transformers/trainer.py` at line 3059 calls `torch.load()` without the `weights_only=True` parameter. This issue affects all versions of the library supporting `torch>=2.2` when used with PyTorch versions below 2.6, as the `safe_globals()` context manager provides no protection in these versions. An attacker can exploit this vulnerability by supplying a malicious checkpoint file, such as `rng_state.pth`, which can execute arbitrary code when loaded. The issue is resolved in version v5.0.0rc3.
A weakness has been identified in huggingface smolagents 1.25.0.dev0. This affects the function evaluate_augassign/evaluate_call/evaluate_with of the file src/smolagents/local_python_executor.py of the component Incomplete Fix CVE-2025-9959. This manipulation causes code injection. It is possible to initiate the attack remotely. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way.
A weakness has been identified in huggingface smolagents 1.24.0. Impacted is the function requests.get/requests.post of the component LocalPythonExecutor. Executing a manipulation can lead to server-side request forgery. It is possible to launch the attack remotely. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way.
A vulnerability in huggingface/text-generation-inference version 3.3.6 allows unauthenticated remote attackers to exploit unbounded external image fetching during input validation in VLM mode. The issue arises when the router scans inputs for Markdown image links and performs a blocking HTTP GET request, reading the entire response body into memory and cloning it before decoding. This behavior can lead to resource exhaustion, including network bandwidth saturation, memory inflation, and CPU overutilization. The vulnerability is triggered even if the request is later rejected for exceeding token limits. The default deployment configuration, which lacks memory usage limits and authentication, exacerbates the impact, potentially crashing the host machine. The issue is resolved in version 3.3.7.
CVE-2025-11844 is an XPath injection vulnerability in Hugging Face's Smolagents version 1.20.0, specifically in the search_item_ctrl_f function. The vulnerability arises from unsanitized user input being directly concatenated into XPath queries, allowing attackers to manipulate query logic. Exploitation can bypass search filters, access unintended DOM elements, and disrupt automated web tasks, potentially leading to information disclosure and manipulation of AI agent interactions. The flaw is fixed in version 1.22.0. The CVSS score is 5.4 (medium severity), with no known exploits in the wild.
