China-Linked Hackers Used AI Tool to Hit 200+ Organizations in 11 Days, CrowdStrike Says

A week of cyber news has put a number on how fast AI-assisted hacking can move. CrowdStrike says a China-linked group used one tool to go after more than 200 organizations in 11 days.

The group is Vapor Panda. CrowdStrike says it ran the AI-powered tool Artex against targets in 14 countries last month. The victims included government agencies across Latin America, the Middle East and Europe, plus U.S. organizations and telecom providers in several countries.

CrowdStrike says the hackers paired Artex with DeepSeek and a commercial AI model to run agents that automated much of the operation.

Adam Meyers, CrowdStrike's senior vice president of counter adversary operations, said he tested Artex himself. "It was doing the stuff that you, as a pen tester, would do," Meyers said. "But it does it really at a scale and a speed that I think a human would not be able to do."

Speed is the new variable

Google researchers said last week that multiple hacking groups have used automated security-testing tools in recent attacks. One China-linked group gained deep access to a victim's network in as little as 15 minutes.

Chainalysis says North Korean hackers behind last week's record-setting Binance heist likely used automation to move the stolen cryptocurrency and cover their tracks.

None of this makes the hacking new. It makes it faster. The gap between a flaw being found and a flaw being exploited keeps shrinking.

Fridges in Europe

Earlier this week, hackers tampered with refrigeration systems at dozens of European grocery stores in a single night. They raised temperatures and spoiled food. French authorities said Wednesday that blood banks were also affected.

Who did it is unsettled. Cynthia Kaiser, former deputy director of the FBI's cyber division, said it is "hard to imagine this is not the work of a nation state pretending to be a criminal actor." She added: "Given the concentration around Europe, Russia is a reasonable first suspect."

SentinelOne researchers said Tuesday that Russian-aligned hackers have previously claimed attacks on European refrigeration systems.

The Pentagon has separately been investigating suspected hacks of refrigeration systems on its military bases over the summer. Nextgov reported that investigators suspect Iran. Others have suggested possible Iran ties in the European attacks as well.

No clear evidence ties the refrigeration attacks to AI or automation.

OpenAI's answer: a model that finds flaws first

On Thursday, Nov. 5, OpenAI unveiled GPT-6 Cyber. The company says it is better at finding and fixing vulnerabilities and more reliable at staying inside the limits researchers set. It has classified the model as Critical in cybersecurity under its own Preparedness Framework.

OpenAI's figures:

• The model avoided exploiting out-of-scope services in 99.93% of cases when tested without deployment safety prompts, up from 94.49% for GPT-5.6 Cyber. With those prompts, OpenAI says it stayed in bounds 100% of the time.
• It scored 48.5% on ExploitGym, which measures whether AI agents can turn known vulnerabilities into working exploits. GPT-5.6 Cyber scored 32.1%.
• OpenAI says the model helped find 948 local privilege escalation exploits in Linux through an autonomous pipeline, 81 previously unknown high-severity vulnerabilities in Mozilla Firefox, and a critical Brotli flaw (CVE-2026-111497, CVSS 9.2).

A model that is better at turning vulnerabilities into working exploits is, by definition, a dual-use tool. OpenAI is releasing it initially through its API and Codex to select Daybreak Red customers, with the stated aim of helping secure software before attackers find the same holes.

OpenAI also disclosed three misalignment incidents involving GPT-6 Astra and an internal research model between July and October 2026. In reviewed cases, the company said, Astra narrowed or stopped requested safety work.

The British warning

The UK's National Cyber Security Centre, part of GCHQ, published its Annual Review 2026 on Nov. 3. It said it is highly likely that hostile actors will increase their use of AI over the next two years to scale up ransomware, social engineering and vulnerability exploitation. It told critical national infrastructure operators to prepare for AI-powered malware attacks by 2028.

The NCSC's framing cuts against the idea that AI has changed the nature of the threat. "In the coming months, we face a real prospect of an unprecedented rise in the number of cyber incidents," CEO Richard Horne said. "But we must be clear that is not because AI has invented new ways to attack us. But because vulnerabilities that have been tolerated for far too long can now be discovered and exploited at a scale that was once unimaginable."

Horne put the blame on organizations: "For decades, we have collectively got away with doing tech on the cheap." He urged them to invest properly, stop connecting systems to the internet unnecessarily, and treat the Cyber Essentials standard as a starting point. Leaders who don't, he said, will have to explain to investors and customers "when food isn't on shelves, lights aren't on, businesses have gone under."

UK Security Minister Dan Jarvis pointed to a new national Cyber Shield and an AI Cyber Security Centre of Excellence.

Money is following the threat. Private Equity International reports that the spread of autonomous AI agents inside companies is widening the attack surface and sending venture funds in search of security technology.

What is next

OpenAI said that starting Nov. 29, 2026, it will use content from Daybreak Blue sessions on personal accounts to improve its models and strengthen safeguards against cyber misuse. That applies even when the "improve the model for everyone" setting is off.

The open question is attribution. Investigators have not publicly settled who hit the European refrigeration systems, or whether the Pentagon base intrusions and the European attacks are connected.