
What happened
vLLM published fixes in 0.31.0 on October 5 and 0.32.0 on October 18 for eight advisories, including GHSA-h3rc-j82p-2c3h: a disguised EPS image can reach Ghostscript and run code, rated High at CVSS 8.1.
Why it matters
That bug lets a crafted image run code as the vLLM process when Ghostscript is present, so teams that carry it in inference images may need to rebuild containers as well as bump the version.
WHO IT HITSPlatform and infrastructure teams running self-hosted vLLM inference containers, including those inside vendor runtimes and air-gapped bundles, need to check their images for Ghostscript and Pillow versions and plan a rebuild.
Summaries like this, every morning.
vLLM's security advisory list now runs to 12 pages, and its first page alone carries ten advisories published between September 23 and October 20, most rated Moderate. A Moderate rating describes one bug in isolation, but on a shared endpoint a single bad request taking out everyone's in-flight work is the outage an on-call team gets paged for.
The Ghostscript path has a second dependency: on a host whose Ghostscript is affected by CVE-2024-29510, a format string flaw in Ghostscript 9.50 through 10.03.0 that bypasses the -dSAFER sandbox, the EPS file can run commands with the vLLM process's privileges. Ghostscript fixed that in 10.03.1 on May 2, 2024. The reporter reproduced command execution at the loader level using the PyPI wheel, Pillow 12.3.0 and Debian's Ghostscript 10.03.0, and traced the full HTTP path in source without testing it end to end.
The October batch differs in reach from earlier vLLM flaws such as CVE-2025-32444, which scored 10.0 for unauthenticated pickle deserialization in the Mooncake integration and hit only teams using one optional component. The hang and the token ID crash need no optional integration. Teams that consume vLLM through a vendor distribution may not be able to tell from the upstream version number whether the fix is in, which is why the cross-tenant crash reporter asked for pre-disclosure notice to LMCache, NIXL and NVIDIA Dynamo, llm-d, vLLM Production Stack and KServe.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.