Home / News / Article
The Record · 4 min read

Chinese state group developing automated attack tool for cyber espionage, Google says

Both criminal and state-backed groups in China are leveraging AI tools for cyber espionage, vulnerability research and social media influence operations.

Google released its latest AI Threat Tracker report on Monday, writing that threat actors are rapidly evolving their use of artificial intelligence in attacks — now leaning more heavily on autonomous multi-agent offensive frameworks like Artex, CAIRN, CyberStrikeAI and PentAGI to execute multi-stage intrusions at greater operational velocity.

While multiple countries — including Russia, Iran and North Korea — are using AI for nefarious purposes, few countries have matched China.

Google researchers said they saw Chinese state-backed and financially motivated actors using AI tools for “reconnaissance, front-end scanning, researching targets across various sectors and geographies, and weaponizing exploits.”

Incident responders saw Chinese contractors using large language models to translate and comb through tranches of stolen data that were eventually organized and sold to customers. Other Chinese actors are using AI platforms’ file-hosting infrastructure to stage archives of phishing lures and exfiltrate stolen data.

Several different Chinese actors are also experimenting heavily with Artex — which was spotlighted on Friday by the FBI in an alert that drew alarm among cybersecurity experts. The tool was created by a Chinese researcher and allows any user to turn a large language model into a powerful hacking tool that can “automate multiple phases of cyber intrusions at machine speed, from reconnaissance through data exfiltration,” according to the FBI.

Google said it found evidence that Chinese government groups are now developing their own version of Artex for “likely use in cyber espionage operations.”

While Artex has garnered the most media coverage, Google said they have seen similar tools used in real attacks or tested by threat actors in private — including CAIRN, PentAGI, CyberStrikeAI and HexStrike AI.

CAIRN allows AI agents to collaborate on breaching targets, often starting with a target’s IP address and ending with an operational objective. Google saw CAIRN used in intrusions of companies in the U.S. aerospace sector.

Some threat actors used PentAGI to conduct “systematic reconnaissance and vulnerability scanning against a wide range of organizations in Asia,” the researchers added.

Vulnerability research and disinformation

Google said multiple countries and vendors now routinely employ large language models to research new bugs, create fake content, manipulate SEO and generate influence content.

The report cites one instance of Iranian government-backed actor APT42 using several large language models, including Google’s Gemini, to conduct targeted reconnaissance and social engineering attacks.

North Korea has continued to deploy AI to support its IT worker campaign, using it to create resumes and supplement knowledge during job interviews.

The researchers added that recent intrusions into cryptocurrency organizations in North America and Europe saw North Korean hackers use AI to create fake recruiter profiles on professional networking profiles.

The fictitious personas then conducted ClickFix attacks — where hackers force victims to take actions on their device to fix problems that are not real. The tactic has been used widely by North Korean actors in multiple attacks on cryptocurrency companies and was recently spotlighted as part of a long-running campaign recently disrupted by the FBI.  

Google found another version of the scheme, with North Korean actors conducting ClickFix attacks and breaching a victim’s device, allowing them to pivot into internal code repositories to install code directly into the large language models the companies use.

The hackers modified code so that when users started up Claude Code, it would automatically trigger the download of a backdoor researchers call SOMBERMEME.

Threat actors are also using LLMs to automate labor- and time-intensive phases of vulnerability analysis, allowing them to quickly analyze patches to find bugs — dramatically reducing the window between patch disclosure and exploitation.

Google said zero-day exploits rose significantly this year, going from eight per month in 2025 to 22 in the month of August.

As an example, Google cited two vulnerabilities that emerged this summer affecting PaperCut print management servers. Researchers were able to generate working exploits for the bug within a day of their disclosure using LLMs.

Researchers noted in their report that while AI tools have compressed attack timelines and allowed some threat actors to move at “machine speed,” they do generate significant forensic traffic and rely on recognizable identifiers, leaving behind distinct code-level traces.

“While these artifacts increase the likelihood of detection and mitigation, these tools remain highly capable of achieving operational objectives,” Google said. “As the sophistication of these tools and underlying LLMs increases, we anticipate not only reduced human involvement in these operations but also increased stealth and sophistication of the activity.”

Originally published on The Record