OpenAI tells macOS users to update ChatGPT, Codex, and Atlas after Axios supply chain scare


MacOS users should update OpenAI’s desktop apps right away. OpenAI says a compromised Axios package reached part of its macOS app-signing workflow, so the company is rotating security certificates and requiring users to move to newer builds.

OpenAI says it found no evidence that user data was accessed, that its systems or intellectual property were compromised, or that its software was altered. The issue affected the process used to certify legitimate macOS apps, not customer accounts or API keys.

The practical takeaway is simple. If you use ChatGPT Desktop, Codex App, Codex CLI, or Atlas on a Mac, install the latest version from the in-app updater or OpenAI’s official download pages. OpenAI says older versions will stop receiving updates or support after May 8, 2026, and they may stop working.

What happened

OpenAI says the incident started on March 31, 2026, when a malicious Axios update got pulled into a GitHub Actions workflow used in its macOS app-signing process. That workflow had access to certificate and notarization material for ChatGPT Desktop, Codex, Codex CLI, and Atlas.

Security researchers say attackers published two backdoored Axios versions, 1.14.1 and 0.30.4. Those packages included a hidden dependency called plain-crypto-js, which triggered a post-install script and fetched platform-specific malware for Windows, macOS, and Linux.

OpenAI says its own analysis suggests the signing certificate was likely not successfully exfiltrated. Even so, the company is treating the certificate as compromised, revoking it, rotating it, and publishing fresh macOS builds as a precaution.

Why the update matters

This update is about trust and software authenticity. A code-signing certificate helps macOS confirm that an app really comes from the claimed developer, and OpenAI says rotating that certificate lowers the risk of fake apps posing as real OpenAI software.

OpenAI also says it has worked with Apple to block new notarization attempts using the previous certificate. That means a fraudulent app signed with the older material should lack notarization and be blocked by macOS by default unless a user manually bypasses security warnings.

For users, the deadline matters almost as much as the incident itself. OpenAI says the earliest safe macOS releases signed with the updated certificate are ChatGPT Desktop 1.2026.051, Codex App 26.406.40811, Codex CLI 0.119.0, and Atlas 1.2026.84.2.

A quick look at the affected apps

AppMinimum version with updated certificateWhat OpenAI says
ChatGPT Desktop1.2026.051Older macOS versions may stop receiving updates or support after May 8, 2026
Codex App26.406.40811Update now through official channels
Codex CLI0.119.0API keys were not affected
Atlas1.2026.84.2Install only from official OpenAI sources

Source for all entries: OpenAI’s incident response page.

Key facts at a glance

  • OpenAI says no user data was accessed.
  • OpenAI says passwords and API keys were not affected.
  • The company says the issue was limited to its macOS app-signing workflow.
  • Researchers linked the broader Axios campaign to malicious npm releases 1.14.1 and 0.30.4.
  • OpenAI says Android, Windows, Linux, and web services were not part of this certificate rotation notice. The action focuses on macOS desktop apps.

FAQ

Do macOS users really need to update now?

Yes. OpenAI says all macOS users of ChatGPT Desktop, Codex App, Codex CLI, and Atlas should update to the latest versions because the company rotated its security certificates.

Was customer data stolen?

OpenAI says it found no evidence that user data was accessed or exposed. Reuters reported the same point in its coverage of the company’s statement.

Do users need to change passwords or API keys?

No. OpenAI says passwords and OpenAI API keys were not affected by this incident.

What happens after May 8, 2026?

OpenAI says older versions of the affected macOS apps will no longer receive updates or support after that date, and they may stop functioning.

How did the broader Axios attack work?

Researchers say attackers published malicious Axios versions that pulled in a fake dependency and then fetched a second-stage RAT tailored to macOS, Windows, or Linux.

Readers help support VPNCentral. We may get a commission if you buy through our links. Tooltip Icon

Read our disclosure page to find out how can you help VPNCentral sustain the editorial team Read more

User forum

0 messages