Reading: Cyberattacks get faster as GTIG, Mandiant flag AI-driven offensive tools

Cyberattacks get faster as GTIG, Mandiant flag AI-driven offensive tools

Published
3 min read
Advertisement

Google Threat Intelligence Group and Mandiant said they have seen a shift in cyberattacks: multiple threat actors are now abusing open source, agentic penetration-testing frameworks to automate parts of intrusions. The move marks a step beyond manual prompting and into autonomous multi-agent offensive frameworks that can compress attack timelines and push operations toward machine speed.

That matters now because the latest observations came last quarter, when investigators saw adversaries using generative AI to scale operations across cyber crime, espionage and influence work. The same shift is showing up in different forms, from automated patch analysis that helps attackers quickly weaponize n-day vulnerabilities to efforts that steal LLM credentials and cloud resources from compromised enterprise environments.

The practical effect is straightforward: threat actors no longer need to keep every step of an intrusion in human hands. The frameworks named by GTIG and Mandiant include ARTEX, CAIRN, CyberStrikeAI and PentAGI, and they are being used to automate discrete phases of operations, including reverse engineering patches through differential analysis. That can turn a newly exposed flaw into a usable exploit pipeline before defenders have finished working through remediation.

- Advertisement -

Financially motivated groups are also using the same tools to resell access to high-performance AI models through underground proxy services, while state-sponsored groups, commercial Influence-as-a-Service vendors and inauthentic political monetization networks use LLMs to generate synthetic media, build localized personas and automate cross-platform distribution. Cyber espionage groups are applying LLMs for target research, automated data analysis and even novel supply chain attack vectors, including cases where autonomous coding assistant configurations are manipulated to drop backdoors during local execution.

There is one caution in the findings. GTIG and Mandiant are drawing on telemetry, incident response engagements, global threat actor tracking and live platform defenses, not a single confirmed campaign, and the report frames the moment as a critical inflection point rather than a finished picture. That makes the unanswered question more important, not less: which specific actors are behind the abuse of these agentic frameworks, and how quickly will defenders adapt once those techniques spread further?

What is already clear is that cyberattacks are becoming easier to automate, harder to attribute and faster to scale. The next test is whether defenders can keep pace with intrusion chains that are increasingly designed to run with less human touch and far more speed.

Advertisement
Share This Article