Hubs › AI and Technology › Post
Join TrustHub to participate — every member is ID-verified
Sign Up Free
▲ 0 ▼

What happens when hackers get AI and defenders don't?

Over the past week a lot of people were convinced the United States was under a cyberattack. Microsoft Outlook and Teams went down. The FAA grounded flights nationwide. Chase online banking broke. Apple had outages, and so did the NSA.

Most of that wasn't hacking. Microsoft said its outage had nothing to do with a cyberattack or AI agents. The Transportation Secretary blamed wiring work for the FAA problem. Chase said its trouble wasn't a cyberattack either, and Apple said an internal update caused theirs. On November 18 the White House put out a statement, and the Director of National Intelligence said it plainly: "This is not a cyberattack."

So the rumor was wrong. But while everyone was staring at the outages, something real was happening, and I think it's a bigger deal.

The FBI is warning about a free AI hacking tool

On November 13 the FBI put out a warning about a tool called ARTEX. It's an open-source "penetration testing" framework, the kind of thing security teams use to attack their own systems and find the holes. A developer put it on GitHub in July. It got popular fast, and criminals started using it.

According to the FBI, attackers have been using ARTEX since at least August. Some break-ins went from first targeting to privileged access in under three hours. Researchers found more than 1,600 ARTEX control panels sitting open on the internet, and the FBI says the real number is probably much higher. The developer took the code down in October, but copies are still everywhere, and there's even a paid version.

Here's the part that matters: ARTEX doesn't have its own brain. You plug in an AI model, and the AI does the hacking work. The FBI says the attackers mostly plug in Chinese models, especially DeepSeek's cheap Flash models, because they're capable, cheap, and don't seem to refuse when the job is framed as security testing.

This isn't theoretical anymore

The FBI lists real victims. In South Korea, attackers using ARTEX stole personal information from about 67,000 people at financial companies. CrowdStrike found the attacker's own servers. DeepSeek was the main model behind ARTEX, and the servers also held Claude Code agent sessions. In a separate supply chain attack on ConnectWise's cloud backup service, attackers tried to sign in to more than 2,250 customer tenants. A tenant is a whole organization's account.

CrowdStrike also tracked a China-linked group it calls Vapor Panda. Over 11 days in October, the group used ARTEX against more than 200 organizations in at least 14 countries and confirmed break-ins at nine of them. A single intrusion used roughly 1.4 billion tokens of AI input. That's not a person typing commands. That's a machine grinding through a network.

Palo Alto Networks reported a separate campaign that used a different open-source tool, called Cairn, with Zhipu's GLM and DeepSeek models. It says, with moderate confidence, that the group behind it is linked to China. The campaign ran 25 operations against at least 19 organizations, including chip supply-chain companies, organizations tied to U.S. defense policy, and civil society groups, and got in at 10 of them.

And on November 19 the European Commission confirmed it was hit. It says the attack was first detected on October 22 and used a publicly available AI tool, but it did not name the tool. Politico and Euronews report that it was ARTEX. Staff systems for payroll, leave and health insurance were reportedly disrupted for more than 10 days. The Commission says it hasn't found evidence of a breach of staff data as a whole.

Meanwhile, regular hacking didn't stop either. Malicious code planted through the Optimizely website tool was stealing card details and crypto. A security firm counted 178 websites carrying it. Optimizely itself confirmed only that some customers were affected. Expedia confirmed a supply chain attack. And Chase said today that some of its outside service providers were hit, and some customers' names and emails may have been exposed.

Why I don't want AI development paused

In my last video I said we can't just stop building AI, because we're going to need our own powerful AI. This is what I meant.

Look at which models the attackers actually used. Mostly cheap Chinese models with no visible guardrails. If American labs paused tomorrow, none of these attackers would lose their tools. The only people who would slow down are the defenders.

And defenders are already using AI, and it's working. In April, Anthropic started a program called Project Glasswing with Apple, Google, Microsoft, CrowdStrike, Palo Alto Networks and others. By May, partners had found more than 10,000 high or critical security holes. Mozilla fixed 271 in a single Firefox release. One partner bank used it to help stop a fraudulent $1.5 million wire transfer.

The government agencies are saying the same thing. In June, the cyber agencies of the US, UK, Canada, Australia and New Zealand put out a joint statement, and their message was blunt: "Defenders must do the same." The head of the UK's cyber agency said defenders need AI "at least as quickly" as attackers use it.

There's a catch, and Anthropic pointed it out itself: AI finds the holes faster than people patch them. Out of an estimated 530 holes it reported to open-source projects, only 75 had been fixed at the time of that update. Finding problems is no longer the slow part. Fixing them is.

Where I land

I think this is unprecedented. I can't think of any other time when hackers in different countries were running automated campaigns against hundreds of organizations at once, with an AI doing most of the work, and a free tool anyone can download.

The outages over the past week were mostly bad luck and bad updates. But the AI attacks are real, they're growing, and they're mostly running on models that nobody here can pause. So the answer can't be to stop. It has to be making sure the people defending hospitals, banks and power grids have AI that's at least as good as what's being used against them.

Do you think a pause would make us safer, or just leave the defenders behind?

Sources

FBI FLASH-20261113-001, "Threat Overview of ARTEX Autonomous Penetration Testing Framework," Nov 13, 2026: https://www.ic3.gov/CSA/2026/261113.pdf

White House, "No Indication of a Cyberattack": DNI Clayton and Secretary Rubio on Recent Tech Outages, Nov 18, 2026: https://www.whitehouse.gov/releases/2026/11/no-indication-of-a-cyberattack-dni-clayton-and-secretary-rubio-on-recent-tech-outages/

Deseret News, officials respond to outage rumors, Nov 18, 2026: https://www.deseret.com/politics/2026/11/18/cyberattack-rumors-government-responds/

AP via ABC7, Apple outage cause, Nov 17, 2026: https://abc7ny.com/story/apple-joins-parade-technology-hiccups-recent-days/20048749/

BleepingComputer, FBI warns ARTEX AI hacking tool is powering global cyberattacks, Nov 15, 2026: https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/

CrowdStrike, ARTEX used against South Korean finance, Oct 7, 2026: https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/

CrowdStrike, Vapor Panda uses ARTEX to scale global campaign, Oct 29, 2026: https://www.crowdstrike.com/en-us/blog/vapor-panda-uses-artex-to-scale-global-campaign/

Palo Alto Networks Unit 42, Cairn framework, Nov 12, 2026: https://unit42.paloaltonetworks.com/cairn-framework-ai-autonomous-attacks/

Euronews, EU Commission hit by large-scale cyberattack, Nov 19, 2026: https://www.euronews.com/2026/11/19/eu-commission-hit-by-large-scale-cyberattack-with-data-leakage-unknown

The Record, European Commission confirms cyberattack, Nov 19, 2026: https://therecord.media/european-commission-confirms-cyberattack-using-ai

The Record, Optimizely malicious code on customer sites, Nov 17, 2026: https://therecord.media/optimizely-malicious-code-customer-sites

Skift, Expedia supply chain attack, Nov 12, 2026: https://skift.com/?p=848794

ABC News, JPMorgan Chase third-party cyberattack, Nov 19, 2026: https://abcnews.com/US/jpmorgan-chase-reports-cyberattack-party-service-providers/story?id=138287705

Anthropic, Project Glasswing, Apr 7, 2026: https://www.anthropic.com/glasswing

Anthropic, Glasswing initial update, May 22, 2026: https://www.anthropic.com/news/glasswing-initial-update

Five Eyes cyber agencies joint statement, Jun 22, 2026: https://www.cisa.gov/news-events/news/five-eyes-cyber-security-agencies-statement

UK NCSC, CEO keynote at CYBERUK, Apr 22, 2026: https://www.ncsc.gov.uk/speech/ncsc-ceo-keynote-speech-cyberuk-2026

0 Comments

Log in to join this hub and comment.

No comments yet. Be the first to reply!