What changed in AI security, Jul 28 to Aug 3, 2025
Jul 28 to Aug 3, 2025 (ISO week 2025-W31). Weeks run Monday to Sunday in UTC.
17 records published, +6 on the previous week: 9 vulnerabilities (-1), 0 incidents (no change), 1 research item (no change), 4 news items (+4), 3 policy items (+3).
Critical and high advisories
Vulnerability records rated critical or high, newest first.- High
CVE-2025-54424: 1Panel remote code execution via incomplete certificate verification
CVE-2025-54424NVD/CVE Database - Critical
CVE-2025-45150: LangChain-ChatGLM-Webui insecure permissions expose sensitive files
CVE-2025-45150NVD/CVE Database - Critical
CVE-2025-50472: modelscope/ms-swift arbitrary code execution via deserialization
CVE-2025-50472NVD/CVE Database - High
CVE-2025-7725: Contest Gallery WordPress plugin stored cross-site scripting via comments
CVE-2025-7725NVD/CVE Database - Critical
CVE-2025-54430: dedupe GitHub workflow runs untrusted PR code on benchmark comment
CVE-2025-54430NVD/CVE Database - Critical
CVE-2025-54381: BentoML server-side request forgery through file upload URL handling
CVE-2025-54381NVD/CVE Database - Critical
CVE-2025-46059: langchain-ai GmailToolkit indirect prompt injection via crafted email
CVE-2025-46059NVD/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-2025-54381: BentoML server-side request forgery through file upload URL handling CVE-2025-54381NVD/CVE Database | Not listed | 16.1% |
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.| Package | Ecosystem | LLM SDKs | Release | Released |
|---|---|---|---|---|
| lfx | PyPI | LangChain, Model Context Protocol SDK | 0.1.0 |
Topics that moved
Largest increases over the mean of the 4 previous weeks, for topics with at least 3 records in the week.| Topic | Records | Weekly mean, previous 4 | Difference |
|---|---|---|---|
| Prompt injection and jailbreaks | 5 | 0.5 | +4.5 |
Research
Peer-reviewed first, then newest.Policy and regulation
Newest first.Generated from the AI Sec Watch database at . Every item links to its record.