Google Threat Intelligence Group Finds Multiple Cyber Attacks Using Pentesting AI — Cyber Attacks

Google Threat Intelligence Group and Mandiant say multiple threat actors are using agentic pentesting frameworks to speed up cyber attacks.

Published
2 Min Read
Google Threat Intelligence Group Finds Multiple Cyber Attacks Using Pentesting AI — Cyber Attacks

Google Threat Intelligence Group and Mandiant say multiple threat actors are now abusing open source, agentic pentesting frameworks to automate phases of cyber attacks. The shift points to faster intrusions with less hands-on work, which can help adversaries move from testing to exploitation before defenders catch up.

- Advertisement -

GTIG AI Threat Tracker

Aurora Blum coauthored the Google Cloud blog update on the GTIG AI Threat Tracker. The update says adversaries are operationalizing generative AI to compress attack timelines, scale operations, and execute multi-stage intrusions at machine speed.

Last quarter, GTIG and Mandiant said adversary tradecraft had already moved beyond manual prompting and into autonomous multi-agent offensive frameworks such as ARTEX, CAIRN, CyberStrikeAI, and PentAGI. That progression means the tooling is no longer limited to one-off prompts, which lowers the amount of human steering needed during an intrusion.

LLMs and n-day exploitation

The same update says threat actors are increasingly leveraging LLMs and agentic frameworks to fully automate the reverse engineering of patches via differential analysis. In practice, that lets attackers turn a freshly patched weakness into a weaponized n-day vulnerability and push exploitation toward internet-facing servers without human intervention.

Google says it is hardening models and disrupting bad actor projects, while the same update says adversaries are already using autonomous frameworks at machine speed. That gap matters because the defenders’ response is moving, but the attacker workflow is already built for scale.

- Advertisement -

LLMJacking and LLM Mixers

The update also says financially motivated threat actors are profiting from LLMJacking by stealing LLM credentials and cloud resources from compromised enterprise environments. Others are reselling access to high-performance AI models through underground proxy services such as LLM Mixers, which turns stolen access into a reusable service instead of a single breach.

Beyond theft and resale, the same update says state-sponsored groups, commercial Influence-as-a-Service vendors, and inauthentic political monetization networks are using LLMs to produce synthetic media at scale, create localized personas, and automate cross-platform distribution. Cyber espionage groups are using LLMs for target research, automated data analysis, and supply chain attack vectors, including manipulating autonomous coding assistant configurations to drop backdoors during local execution.

What GTIG and Mandiant still do not spell out is how many real-world intrusions were enabled by these agentic frameworks. That leaves the size of the shift open, even as the tooling itself has clearly moved from assisted prompting to autonomous execution.

Advertisement
Share This Article
Tech writer covering AI, cloud infrastructure, and enterprise software. Former software engineer at Google with 7 years in technology journalism.