LlamaIndex vulnerable to remote code execution in JsonPickleSerializer
llama_index
An insecure fallback to Python's pickle module in LlamaIndex allows remote code execution via crafted payloads. Attackers can leverage this to achieve full system compromise.
disclosed CVENo confirmed exploitation, so this record is tracked as a disclosed vulnerability and not counted as an incident. It moves into the counts if CISA or the vendor confirms exploitation. Why
- CVSS
- 5.0
- EPSS
- 0.01 · top 57% · 2026-11-13
- fix status
- Fix status unknown
- first report
- 7 Jul 2025
- last update
- 7 Jul 2025
- CVE ids
Timeline · 1
Sources · 1
Advisories and databases
- 2025-07-07CVE-2025-3108: LlamaIndex has Incomplete Documentation of Program Execution related to JsonPickleSerializer componentGitHub Advisory Database · github.com/advisories/GHSA-m84c-4c34-28gf
Titles and links belong to their publishers. The summary above is written by gcdRadar in its own words; data is CC BY-SA 4.0.
Classification
- lane
- Patch now
- AI role
- Attack on AI
- categories
- Infrastructure vulnerability
- severity
- 2 · Disclosed vulnerability
- fix status
- Fix status unknown
- targets
- Framework / tooling
- owasp llm top 10
- LLM03:2025 Supply Chain
- mitre atlas
- no technique for these categories
- dates
- happened 7 Jul 2025 · first reported 7 Jul 2025
- origin
- GitHub Advisory DatabaseVulnerability record from NVD / the GitHub Advisory Database. This product uses the NVD API but is not endorsed or certified by the NVD.
- record
- confidence 0.95 · created 2026-10-18 07:00Z · updated 2026-10-18 07:00Z
Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned
Copy a brief
This record as a short Markdown brief, for a ticket, a chat or a coding agent.
the brief as text
# LlamaIndex vulnerable to remote code execution in JsonPickleSerializer https://gcdradar.vercel.app/incidents/llamaindex-vulnerable-to-remote-code-execution-in-jsonpickle · AIR-2025-0311 · closed 2025-07-07 - Lane: Patch now - Severity: 2 (Disclosed vulnerability) - Fix status: Fix status unknown - First reported: 2025-07-07 - AI role: Attack on AI - Products: llama_index - CVE ids: CVE-2025-3108 - CVSS: 5 ## Summary An insecure fallback to Python's pickle module in LlamaIndex allows remote code execution via crafted payloads. Attackers can leverage this to achieve full system compromise. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2025-07-07: CVE-2025-3108: LlamaIndex has Incomplete Documentation of Program Execution related to JsonPickleSerializer component (https://github.com/advisories/GHSA-m84c-4c34-28gf) ## Sources - 2025-07-07 GitHub Advisory Database: CVE-2025-3108: LlamaIndex has Incomplete Documentation of Program Execution related to JsonPickleSerializer component https://github.com/advisories/GHSA-m84c-4c34-28gf Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- closed6.5LlamaIndex LlamaIndex vulnerable to DoS through recursive JSON parsing
Patch now · Disclosed · since 7 Jul 2025 · CVE-2025-5472
AI - closed7.5LlamaIndex CVE-2024-12704: LlamaIndex denial of service vulnerability
Patch now · Disclosed · since 20 Mar 2025 · CVE-2024-12704
AI - closed8.8LlamaIndex CVE-2024-4181: LlamaIndex RunGptLLM Command Injection
Patch now · Disclosed · since 16 May 2024 · CVE-2024-4181
AI - closed6.3Agno CVE-2025-8665: OS command injection in Agno MCP Tools
Patch now · Disclosed · since 6 Aug 2025 · CVE-2025-8665
AI - closed5.3Hugging Face libraries CVE-2025-5197: Hugging Face Transformers ReDoS Vulnerability
Patch now · Disclosed · since 6 Aug 2025 · CVE-2025-5197
AI - closed8.7Claude Code + SDKs CVE-2025-54795: Claude Code Command Prompt Execution Bypass
Patch now · Disclosed · since 4 Aug 2025 · updated 5 Aug 2025 · CVE-2025-54795
AI