Despite several vulnerabilities being mentioned in the report, only one is explicitly referenced: CVE-2026-66384, which is attributed to OpenAI's Kostya Kortchinsky. Going through JFrog's security advisories, there are at least 20 other CVEs by OpenAI in the last few months, including the types described in the report: SSRF, privilege escalation, arbitrary file write, etc. Some of them are: CVE-2026-70551, CVE-2026-42016, CVE-2026-66014, CVE-2026-66015, CVE-2026-65616, CVE-66375, CVE-2026-69104, CVE-2026-69105.
Post summary
The post enumerates multiple OpenAI CVEs (e.g., CVE-2026-66384) and notes their categories (SSRF, privilege escalation, etc.), but provides no PoC, exploit code, active exploitation evidence, patches, or debunking claims, serving as a general inventory.