Research explained

How o3 helped a researcher find a Linux SMB memory-safety flaw

The vulnerability affected Linux ksmbd; OpenAI’s model was used during discovery.

Published 29 Oct 2026 · Material update 26 May 2025 — last time the underlying evidence changed · Evidence: high

What happened

In May 2025, researcher Sean Heelan described using o3 during an investigation of Linux’s ksmbd implementation. The work uncovered CVE-2025-37899, a use-after-free involving the SMB logoff handler and concurrent connections.

The affected software is the Linux SMB implementation. The involvement of an AI model in finding the bug does not make OpenAI the vulnerable vendor.

Heelan’s account is useful because it describes the evaluation setup, code context, and limitations rather than only announcing a successful result. For teams evaluating AI-assisted research, retain that distinction: a valid finding is evidence of capability, not a guarantee that every run will be accurate.

Operators should consult their Linux distributor’s advisory for package applicability and fixes. This retrospective does not determine whether a specific host is exposed.

Read the original source.

What we know

Affected component
Linux ksmbd
AI role
Discovery assistance

What remains unknown

Open question
The sources reviewed do not establish whether your deployment was exploited.

Source references

Technical details

CVE
CVE-2025-37899