theaisafemode.com
AI Safe Mode

Hijacked AI Tool Update Stole Passwords and AI Keys

Photo of hands typing on a laptop with code open on the screen
Photo: Crew via Unsplash / Wikimedia Commons (CC0 1.0). Illustrative image only.

A hacked update of a popular open-source AI gateway tool carried malware that stole passwords, cloud logins and AI keys. Who is at risk and what to do now.

Updating software is usually the safe thing to do. This week, one update of a popular AI tool was the attack itself.

According to BleepingComputer (Oct 22, 2026) and The Register (Oct 21, 2026), attackers took over the release process of a widely used open-source AI gateway, a tool that connects apps to many AI services at once, and published a version carrying password-stealing malware.[1] [2]

What happened

Both outlets report the malicious version went out on Oct 21 through the tool’s normal download channels, including its official container images.[1] [2] The Register reports the tool gets about 23 million downloads a week.[2]

The malware added hidden files that run every time the programming environment starts, so a developer did not even need to use the AI tool for it to fire.[1] [2] BleepingComputer reports that even running basic commands to investigate a machine could trigger it.[1]

According to both outlets, it went after cloud logins, AI provider keys, SSH keys, saved browser passwords and cookies, and environment variables.[1] [2] It encrypted what it found and sent it to a web address controlled by the attackers, and also installed a backdoor that checks in with the attackers every hour for new commands.[1] [2]

The bad release was removed within hours.[1] [2] The Hacker News also reported the attack as a supply chain compromise delivering a credential stealer and a backdoor.[3]

Why it matters

Tools like this sit in the middle of everything. They hold the keys to several AI services, plus the cloud accounts they run on. One poisoned update can hand all of it over at once.

It is also not the first time. BleepingComputer reports the same project suffered an earlier malicious release in March.[1] Attackers know AI tools are where the valuable keys are.

And because it arrived as an official update, normal good habits, like staying up to date, were what let it in.

How to stay safe

  • Check if you were exposed. If your team installed or ran the Oct 21 release, treat those machines as compromised.
  • Rebuild, don’t just uninstall. BleepingComputer warns that removing the package alone may leave the backdoor running, and The Hacker News advises replacing exposed keys from a clean device. Restore from a clean backup.
  • Replace every secret. Change the passwords, cloud keys and AI keys those machines could reach, from a clean device.
  • Pin versions and wait. In important systems, install exact versions you have checked, and give brand-new releases a short delay.
  • Cap AI spending. A spending limit on every AI account limits what a stolen key can cost you.

The bottom line

An official update can still be poisoned. Keep AI keys in as few places as possible, cap what they can spend, and have a plan to replace them fast.

I don’t hate AI. I just want it safe.

This article is based only on the published news reports listed below. Dates are as published by each outlet.

FAQ

What was stolen?

BleepingComputer and The Register report the malware targeted cloud logins, AI provider keys, SSH keys, saved browser passwords and cookies, and environment variables.

Is uninstalling the bad version enough?

No. BleepingComputer warns that removing the package alone may leave startup files or a running backdoor behind, and The Hacker News advises replacing exposed keys from a clean device.

How long was the bad version available?

The Register reports it was quarantined within hours of being uploaded on Oct 21.

See the post on YouTube Watch and follow AI Safe Mode on YouTube @AISafeMode and Instagram @aisafemode.

References

  1. BleepingComputer, Oct 22, 2026. AI gateway release pipeline hacked to distribute info-stealer malware. Read the report ↗
  2. The Register, Oct 21, 2026. AI gateway supply chain attack hoovers up passwords and secrets. Read the report ↗
  3. The Hacker News, Oct 21, 2026. AI gateway compromised in supply chain attack delivering credential stealer and backdoor. Read the report ↗
Facts come only from the published reports above and are attributed to them. AI Safe Mode doesn't name companies, products or people involved in incidents. Spotted an error? Email [email protected] and it will be reviewed and corrected. See the corrections policy.