openaienmodel: gpt-5-mini-2025-08-07
Our response to the TanStack npm supply chain attack
Key Points
- Certificates rotated; macOS updates required by 2026-06-12
- No evidence of customer data or product compromise
- CI/CD and package-provenance controls accelerated
Summary
On May 11, 2026 OpenAI identified a supply chain compromise of the TanStack npm ecosystem (part of the Mini Shai-Hulud campaign). Two employee devices were impacted; limited credential material was exfiltrated from a small set of internal repositories. There is no evidence of customer data access, production compromise, or malicious signing of released software. As a precaution we are rotating code-signing certificates and coordinating with platform providers to block notarization with the impacted keys.
Key Points
- Incident: TanStack npm compromise affected two employee devices and a limited set of internal repos; DFIR engaged and containment executed.
- Impact: No evidence of user data or product compromise; limited credential material exfiltrated from specific source repos that contained signing certificates.
- Certificates and macOS: We are rotating code-signing certificates; macOS users must update affected OpenAI apps by June 12, 2026 to continue running launches and downloads.
- Platforms: Windows and iOS do not require user action; macOS apps affected include ChatGPT Desktop, Codex App, Codex CLI, and Atlas.
- Mitigations already taken: isolation of impacted systems, session revocation, credential rotation, temporary deployment restrictions, notarization blocks, and full review of prior notarizations and published builds.
- Longer-term controls: strengthened CI/CD credential hardening, package manager controls (e.g., minimumReleaseAge), provenance validation, and additional supply-chain protections.
Recommended actions for engineers
- macOS users: update OpenAI apps via in-app update or official download links before 2026-06-12.
- Verify build/signing pipelines: rotate any exposed keys, confirm CI/CD credential vaulting and access controls, and validate that notarization is blocked for old certs.
- Strengthen dependency provenance: enforce package manager controls, enable provenance validation, and monitor for anomalous package changes.
- Audit access and secrets: review repository access, rotate credentials where appropriate, and continue DFIR collaboration if suspicious activity is detected.
Where to get updates
- Use only official in-app updates or OpenAI download pages for ChatGPT Desktop, Codex App, Codex CLI, and Atlas. Do not install from third-party links.