AI
Google says attackers used AI agents to steal credentials in under six hours
The Google Threat Intelligence Group released a report detailing how threat actors used a multi-agent AI framework to compromise thousands of credentials
Key takeaways
- Threat actors used a multi-agent AI framework to compromise thousands of credentials in under six hours.
- The autonomous framework was assembled from an AI coding chatbot, a prompt, and agent instructions, running troubleshooting and IP rotation without an operator.
- A China-linked espionage group attempted to use Gemini to design an automated penetration testing framework, but Google disabled the assets.
- The criminal group UNC6780 (TeamPCP) has run compromises across PyPI, npm, and Docker Hub, dropping credential stealers into hidden directories like .claude and .cursor.
- Every threat actor is now assumed to be using AI in some capacity and benefiting from its speed.
The Google Threat Intelligence Group released a report detailing how threat actors used a multi-agent AI framework to compromise thousands of credentials in under six hours. Mandiant investigators traced the campaign to a suspected financially motivated actor that built an autonomous framework using an AI coding chatbot, a prompt, and agent instructions. The report highlights a shift toward reducing human involvement in cyberattacks, shrinking the response window for defenders.
Autonomous AI agents are dramatically shrinking the window defenders have to react to cyberattacks.
In their words
“Criminals, like the ones who conducted a mass exploitation campaign in just six hours, will gravitate to attacks that are faster than we can respond to”
By the numbers
- 6 hours
- Time taken by attackers to compromise thousands of credentials
- Q2
- Quarter of activity tracked in the GTIG report
How it unfolded
- UNC6780 poisoned the LiteLLM gateway
- Attacker provisioned GPU instances via exposed GitHub token
- GTIG report documented first AI-built working zero-day exploit
- Google Threat Intelligence Group released its adversarial AI report
Turn stories like this into views
Ravenclip finds the AI news, makes the video, and posts it before attention moves on.
Common questions
- What happened with Threat actors?
- Threat actors used a multi-agent AI framework to compromise thousands of credentials in under six hours.
- Where can I read the original report?
- Read the full report at Siliconangle.
