Hackers Turned an AI Pentest Agent Into a Key Thief

Attackers reportedly used an open-source AI hacking agent, built for security testing, to break into servers and steal cloud keys and production secrets.
Security researchers say attackers used an open-source AI agent made for penetration testing, the authorised hacking companies pay for to find their own weak spots, to break into real systems and steal cloud keys and production secrets.[1] [2]
What happened
According to the reports, the attackers used the AI agent to exploit weaknesses in web applications on servers, then planted web shells, small hidden programs that give remote control.[1] [2]
From there they reportedly collected credentials stored in configuration files and used valid cloud keys to explore the victims’ cloud accounts. The reports describe hundreds of cloud requests within about 30 minutes, and some stored production secrets were retrieved.[1] [2]
Why it matters
Security tools have always been usable by both sides. What AI agents add is speed and automation: one person can run many steps that once needed a skilled team, around the clock.
The weak points were familiar ones. Unpatched web apps, passwords sitting in config files and keys with too much access turned a single break-in into a cloud-wide problem. AI did not create those gaps. It moved through them quickly.
The short version
An AI agent built to test defences was turned into an attack tool. The fix is the same as before, done more strictly and more quickly.
How to stay safe
- Patch internet-facing apps first. Prioritise anything reachable from the internet.
- Keep secrets out of config files. Use a proper secrets store, and never leave keys in code or settings files.
- Limit what each key can do. Give each cloud key the smallest set of permissions it needs.
- Rotate keys regularly. Short-lived keys reduce the damage if one is stolen.
- Alert on unusual activity. Watch for bursts of cloud requests and unexpected reads of stored secrets.
Follow AI Safe Mode
Protect yourself with Privaku
- Privaku Photos (iPhone) keeps your private photos safe: https://apps.apple.com/app/id6821608477
- Privaku AI Shield (Mac) watches what AI agents do on your computer: https://privaku.com/agent-shield/
See the post on YouTube Watch and follow AI Safe Mode on YouTube @AISafeMode and Instagram @aisafemode.
References
- The Hacker News, Nov 11, 2026. Report on attacks linked to an AI pentesting platform. Read the report ↗
- BleepingComputer, Nov 11, 2026. Report on the AI agents used in credential-theft attacks. Read the report ↗
