The report says the issue affected llm’s extra-openai-models.yaml mechanism, now managed through the llm openai endpoint add command. That mechanism supports the use of OpenAI-compatible API endpoints from other providers. According to the report, the problem created two possible exposure paths for a configured OpenAI API key.
One path occurred when the key identified by apikeyname had never been stored, while the other occurred when a model configured with responses: true was run with -o chatcompletions 1. Simon Willison’s Weblog links the report to GitHub advisories GHSA-v9h7-jj4g-f862 and GHSA-r639-rr7r-x72w.
The supplied report does not provide severity ratings, affected llm versions or detailed remediation guidance for the advisories. It was published on November 11, 2026, but the supplied material does not independently establish a separate software release timestamp or provide corroboration from another source.