Vulnerabilities
All
My stack
Known exploited
Public exploit
Critical (9+)
Likely exploited (EPSS ≥ 50%)
Changed, 24 h
New, 24 h
Sort:
risk
modified
published
CVSS
EPSS
| CVE | CVSS | EPSS | Vendor, product | Description | Modified |
|---|---|---|---|---|---|
| CVE-2026-58116exploit | 9.3 critical | 1.0% | hiyouga, llama-factory | LLaMA-Factory through 0.9.5 contains a remote code execution vulnerability that allows attackers with WebUI access to execute arbitrary Python code by supplying a malicious model path in the Chat or Training interfaces. The application passes user-supplied model path input unvalidated into AutoTokenizer.from_pretrained() and AutoModel.from_pretrained() with a hardcoded trust_remote_code=True parameter, causing the Hugging Face transformers library to fetch and execute arbitrary code from a remote or local model repository with the privileges of the server process. | 25 Oct |
| CVE-2026-47117exploit | 9.3 critical | 1.0% | no CPE yet | OpenMed before 1.5.2 contains a remote code execution vulnerability in the PII privacy-filter model loading path. The privacy-filter dispatcher used broad substring matching on the user-supplied model_name parameter, allowing a value such as attacker/foo-privacy-filter-bar to route through a path that loads Hugging Face models with trust_remote_code=True. An unauthenticated attacker can supply a malicious model repository containing custom Transformers code via auto_map in config.json or tokenizer_config.json, which is imported and executed with the privileges of the OpenMed service process. | 8 Oct |
| CVE-2026-61539 | 10.0 critical | 1.0% | no CPE yet | Xinference is an inference API for running open-source, speech, and multimodal models. In 2.5.0 and earlier, Xinference passes attacker-influenced Llama3 tool-call output to eval() in xinference/model/llm/tool_parsers/llama3_tool_parser.py and xinference/model/llm/utils.py. Requests to /v1/chat/completions with a tools field flow through xinference/api/restful_api.py, xinference/model/llm/transformers/core.py, handle_chat_result_non_streaming(), and _post_process_completion() before extract_tool_calls() or _eval_llama3_chat_arguments() evaluates the model-generated Python expression. An unauthenticated remote attacker can influence that output through a crafted prompt and execute commands in the Xinference server process context. This issue is fixed in version 2.7.0. | 12 Oct |
| CVE-2026-14535 | 9.8 critical | 1.0% | trailofbits, fickling | In Trail of Bits fickling versions up to and including 0.1.11, the UnsafeImportsML analysis pass unconditionally calls AnalysisContext.shorten_code(node) on every import node it inspects, regardless of whether the import is flagged as unsafe. This call registers the shortened code representation in the shared AnalysisContext.reported_shortened_code set. When the MLAllowlist analysis pass subsequently runs, it calls the same shorten_code() method, receives already_reported=True for every import, and executes a continue statement that skips its allowlist check entirely. This renders MLAllowlist dead code for all imports — it never evaluates whether an import is in the ML allowlist or not. The MLAllowlist pass was designed to catch imports of modules outside the known-safe ML ecosystem (torch, numpy, transformers, etc.) that slip past the UnsafeImports denylist. With MLAllowlist inoperative, any standard library module not in the UNSAFE_IMPORTS denylist can be invoked via pickle deserialization while fickling's check_safety() returns LIKELY_SAFE. The fickling.load() API chains check_safety() into pickle.loads() as an explicit security gate, meaning a LIKELY_SAFE verdict causes the payload to be deserialized and executed. The root cause is shared mutable state between independently-correct analysis passes — UnsafeImportsML works as designed in isolation, MLAllowlist works as designed in isolation, but the shared reported_shortened_code set causes UnsafeImportsML to poison MLAllowlist's deduplication logic. | 25 Oct |
| CVE-2026-32311 | 9.3 critical | 1.0% | flowsint | Flowsint is an open-source OSINT graph exploration tool designed for cybersecurity investigation, transparency, and verification. Flowsint allows a user to create investigations, which are used to manage sketches and analyses. Sketches have controllable graphs, which are comprised of nodes and relationships. The sketches contain information on an OSINT target (usernames, websites, etc) within these nodes and relationships. The nodes can have automated processes execute on them called 'transformers'. A remote attacker can create a sketch, then trigger the 'org_to_asn' transform on an organization node to execute arbitrary OS commands as root on the host machine via shell metacharacters and a docker container escape. Commit b52cbbb904c8013b74308d58af88bc7dbb1b055c appears to remove the code that causes this issue. | Mon 26 Oct |
| CVE-2026-68770 | 9.3 critical | 1.0% | no CPE yet | sentence-transformers contains a security control bypass vulnerability that allows attackers to achieve arbitrary code execution by exploiting a logic flaw in the import_module_class helper within sentence_transformers/util/misc.py, where the guard condition includes an 'or os.path.exists(model_name_or_path)' clause that satisfies the trust gate whenever the supplied path exists on the local filesystem, regardless of the trust_remote_code=False argument. Attackers who can control or influence the contents of a model directory on disk can place malicious Python files such as modeling_*.py referenced via modules.json, causing the code to execute at import time when an application loads the model with SentenceTransformer(path, trust_remote_code=False), bypassing the documented security contract and achieving code execution within the loading process. | 23 Oct |
| CVE-2026-114053 | 9.2 critical | 0.00% | no CPE yet | vLLM from 0.7.3 before 0.31.0 forwards the request-supplied mm_processor_kwargs.code_revision to Transformers AutoProcessor.from_pretrained, letting clients select which revision of a model's dynamic processor code is imported when trust_remote_code is enabled. An attacker controlling a revision in the served model repository can send a multimodal request to the unauthenticated /tokenize endpoint to execute arbitrary Python in the vLLM API process. | Tue 27 Oct |
7 of 7
Download these 7 as CSV · subscribe to this list.
EPSS scores: 2026-10-31. NVD cursor: 2026-10-29 00:22. The News page shows only notable CVEs (KEV, your stack, new criticals, EPSS over 50%, or active exploitation per SSVC); this table has all of them.