Sunday, November 22, 2026 Washington. Times are Eastern.

Agencies

Google Warns of Autonomous AI Hacking Tools in Aerospace Intrusion

Google's Threat Intelligence Group says attackers used the open-source Cairn framework in a U.S. aerospace intrusion, and the FBI warns about Artex.

Googleplex headquarters building in Mountain View, California.
Google's Mountain View headquarters, the Googleplex, shown in a general exterior view. Photograph: Asoundd/Wikimedia Commons (CC BY-SA 4.0)

Hackers used the open-source Cairn framework in an intrusion targeting the U.S. aerospace sector, Google's Threat Intelligence Group said in a report released Monday, Nextgov/FCW reported. Google researchers said adversaries have moved beyond manual prompting to deploying autonomous multi-agent offensive frameworks that "execute multi-stage intrusions at greater operational velocity," according to Cybersecurity Dive. Google said in a statement released with the report: "The era of autonomous attacks has arrived."

The report left unnamed the aerospace organization that was accessed, did not state what data may have been taken, and did not blame any particular government. Cairn is a penetration-testing framework that is publicly available and has connections to the Chinese open-source security community; it supports AI agents that exchange information and choose hacking methods. The intrusion's timing is unclear, though the report broadly reviews activity from "the last quarter."

The FBI warned on Friday that the Artex framework poses a serious threat, reducing the time required to chain attack techniques, Cybersecurity Dive reported. The bureau said some Artex-related intrusions moved from first targeting to privileged access in less than three hours. Google also observed activity involving CyberStrikeAI, HexStrike AI and PentAGI, all open-source frameworks that researchers said are being used to "execute multi-stage intrusions at greater operational velocity."

Google researchers located an exposed server running an AI-driven hacking tool called exploit_agent, Newsweek reported. The software was built to review updates that developers put out, figure out which security flaws had been patched, and craft attacks aimed at computers that had not yet installed those updates. It could then try to exploit those weaknesses without a person directing each step. The server also contained stolen data and utilities for searching compromised networks, accessing accounts, and installing malware. Evidence suggested the operator was interested in French websites in the media, entertainment and education sectors.

Chinese hacking groups have used AI models to read seized documents and prepare intelligence reports, while North Korean hackers have used fake recruiter profiles and tampered AI coding tools to enter cryptocurrency companies, according to Google's findings. Attackers sometimes broke into companies' cloud accounts and used their computing power to run AI systems, resulting in charges of more than $100,000 in one day and up to $1 million in the worst cases, Newsweek reported. Google said that groups tied to Russia, China and Iran now commonly employ models to generate content and arrange its spread.

John Hultquist, chief analyst for Google's Threat Intelligence Group, said attackers are not using new methods but "are still going after known, unfixed issues, just with much greater speed and scale." He warned that security teams will have to lean more on automated defenses. Google anticipates that the technology will get better, making attacks more difficult to detect. The company is also building its own AI-powered security tools to find and fix vulnerabilities before criminals can exploit them.

Sources: nextgov.com, Newsweek, Cybersecurity Dive

Malik Osei Oversight correspondent

Malik Osei covers inspectors general, the Government Accountability Office and federal spending for West Wing Times. Malik writes from audit reports, budget documents and published reporting, and links every source.

Corrections. Report an error to corrections@westwingtimes.com. We list substantive corrections on the corrections page.