What changed in AI security, Dec 22 to Dec 28, 2025
Dec 22 to Dec 28, 2025 (ISO week 2025-W52). Weeks run Monday to Sunday in UTC.
24 records published, +10 on the previous week: 15 vulnerabilities (+10), 0 incidents (no change), 9 research items (+2), 0 news items (-1), 0 policy items (-1).
Critical and high advisories
Vulnerability records rated critical or high, newest first.- High
GHSA-rcfx-77hg-w2wv: FastMCP updated to MCP 1.23+ due to CVE-2025-66416
GitHub Advisory Database - High
CVE-2025-67729: LMDeploy insecure deserialization when loading model checkpoint files
CVE-2025-67729NVD/CVE Database - High
CVE-2025-68665: LangChain JS serialization injection through toJSON and JSON.stringify
CVE-2025-68665NVD/CVE Database - Critical
CVE-2025-68664: LangChain serialization injection in dumps and dumpd functions
CVE-2025-68664NVD/CVE Database - High
CVE-2025-14930: Hugging Face Transformers GLM4 deserialization of untrusted data
CVE-2025-14930NVD/CVE Database - High
CVE-2025-14929: Hugging Face Transformers deserialization of untrusted data in X-CLIP
CVE-2025-14929NVD/CVE Database - High
CVE-2025-14928: Hugging Face Transformers HuBERT convert_config code injection
CVE-2025-14928NVD/CVE Database - High
CVE-2025-14927: Hugging Face Transformers code injection in SEW-D convert_config
CVE-2025-14927NVD/CVE Database - High
CVE-2025-14926: Hugging Face Transformers code injection in SEW convert_config
CVE-2025-14926NVD/CVE Database - High
CVE-2025-14924: Hugging Face Transformers megatron_gpt2 deserialization flaw
CVE-2025-14924NVD/CVE Database - High
CVE-2025-14921: Hugging Face Transformers code execution via deserialization of model files
CVE-2025-14921NVD/CVE Database - High
CVE-2025-14920: Hugging Face Transformers Perceiver deserialization leads to code execution
CVE-2025-14920NVD/CVE Database - High
CVE-2025-13707: Tencent HunyuanDiT deserialization flaw in model_resume enables code execution
CVE-2025-13707NVD/CVE Database - Critical
GHSA-q9r5-6hrr-9ph7: Hugging Face smolagents: Unsafe deserialization in Remote Python Executor leads to RCE
CVE-2025-14931GitHub Advisory Database - High
CVE-2025-63664: GT Edge AI Platform conversation messages API exposes other users' history
CVE-2025-63664NVD/CVE Database
Exploitation signals
Vulnerabilities published in the week that are listed in the CISA Known Exploited Vulnerabilities catalog or have an EPSS score of 10% or more.| Advisory | Exploitation | EPSS | Published |
|---|---|---|---|
| CVE-2025-68664: LangChain serialization injection in dumps and dumpd functions CVE-2025-68664NVD/CVE Database | Not listed | 42.9% |
Packages that began delegating to a language model
Exposure Registry packages whose first release declaring an LLM SDK, agent framework or MCP dependency was published in the week.No tracked package published its first release with an LLM SDK, agent framework or MCP dependency in this week.
Topics that moved
Largest increases over the mean of the 4 previous weeks, for topics with at least 3 records in the week.No topic had at least 3 records in this week and more than its mean over the 4 previous weeks.
Research
Peer-reviewed first, then newest. Showing 8 of 9.HGNN Shield: Defending Hypergraph Neural Networks Against High-Order Structure Attack
Peer-reviewedIEEE Xplore (Security & AI Journals)Neural Machine Unranking
Peer-reviewedIEEE Xplore (Security & AI Journals)Generative Artificial Intelligence: Ethical Challenges and Trust Mechanisms
Peer-reviewedIEEE Xplore (Security & AI Journals)Large Language Models in Human Subject Research, and the Presence of Idiosyncratic Human Behaviors
Peer-reviewedIEEE Xplore (Security & AI Journals)Enhanced Masking-Differential Prompting (MDP): Defending Backdoor Attacks for Pre-trained Language Models Under Few-Shot Learning
Peer-reviewedIEEE Xplore (Security & AI Journals)Slack Federated Adversarial Training
Peer-reviewedIEEE Xplore (Security & AI Journals)Proactive Bot Detection Based on Structural Information Principles
Peer-reviewedIEEE Xplore (Security & AI Journals)Exploring the Vulnerabilities of Federated Learning: A Deep Dive Into Gradient Inversion Attacks
Peer-reviewedIEEE Xplore (Security & AI Journals)
Policy and regulation
Newest first.No regulatory or policy records were published in this week.
Generated from the AI Sec Watch database at . Every item links to its record.