CrowdStrike says agentic AI helped Vapor Panda scale cyberattacks across 14 countries

A China-backed group tracked by CrowdStrike as Vapor Panda used an agentic penetration-testing tool in a campaign targeting more than 200 organisations across at least 14 countries, CyberScoop reports. CrowdStrike says about 140 sensitive targets were compromised in 11 days, a striking example of AI accelerating familiar cyberattacks rather than conjuring a new kind of exploit.

Watch Desk analysis

What happened

CrowdStrike says the group used ARTEX, an automated agentic AI penetration-testing framework, alongside DeepSeek. Its targets included government, defence and intelligence bodies, as well as US organisations in sectors including pharmaceuticals, semiconductors, nuclear and finance. The campaign remains active, according to the report.

CrowdStrike told CyberScoop that roughly 70% of targets were confirmed compromised or had evidence indicating a moderately likely intrusion. The company also says ARTEX activity reached more than 1,500 commands an hour, compared with up to 40 commands an hour in the hands-on-keyboard activity it tracked over the past 15 years. Those figures and the attribution are CrowdStrike’s, as reported by CyberScoop.

Why it matters

The unsettling change is pace and scale. Automating parts of an attack can let operators pursue multiple paths across many targets at once, while the reported activity still relies on familiar techniques. Organisations cannot treat “AI-powered” as a synonym for a novel exploit, or assume ordinary security basics have stopped mattering.

Our read

This is a significant warning about how agentic tools may amplify an attacker’s reach, not evidence that the software is conducting fully independent campaigns. CrowdStrike’s practical advice, reported by CyberScoop, is to maintain visibility across endpoints, identities and systems. Less glamorous than an AI arms race, perhaps, but rather more useful at 3am.

What to watch

• Whether CrowdStrike provides further detail on the campaign’s targets and confirmed intrusions.
• Whether other security researchers observe similar use of agentic tools at this scale.
• How defenders adapt monitoring as attack activity becomes faster and more automated.

Discussion spark: If AI mainly makes familiar attacks faster and broader, should defenders prioritise better automation or stronger identity and endpoint controls?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.