“Hackers are now using AI agents that can work around the clock and try many ways in at once, and this week some of the biggest names in security showed it happening in real attacks. Microsoft says the AI framework used by the China-linked group, called ARTEX, is also being used by attackers in other parts of the world, while CrowdStrike…”

ARTEX and Microsoft. Microsoft's October 21 blog states that as early as September 2026, Storm-0940, a China-state aligned actor, used the agentic AI penetration testing framework ARTEX to automate attacks targeting hundreds of organizations globally across a broad range of industries. On the specific question of use beyond this one group, Microsoft writes plainly: "Microsoft Threat Intelligence has identified threat actors across multiple geographic regions leveraging the ARTEX framework." Microsoft also characterises agentic tooling in the terms the post paraphrases: agents are not constrained by factors affecting humans, they can work around the clock, on weekends, and on holidays and attempt multiple breach paths simultaneously. On scale, Microsoft observed more than 151,000 failed logon attempts during the campaign, peaking at 2,900 failures per minute, and Storm-0940 used ARTEX to reach more than 2,250 tenants. Separately and earlier, CrowdStrike Intelligence identified a campaign against South Korean financial organizations active from late September to early October 2026, in which the threat actor leveraged ARTEX, a recently released open-source agentic pentesting tool developed in China, alongside large language models, a campaign Zscaler ThreatLabz independently analysed, reporting exfiltration of personally identifiable information belonging to approximately 67,000 individuals. CrowdStrike and the four actors. CrowdStrike's own report page states the finding in close to the post's words: it uncovered four previously unreported threat actors weaponizing AI agents in real-world intrusion campaigns, and under the heading that AI safeguards can be bypassed, states that all four threat actors overcame commercial AI safety controls to enable malicious activity. CrowdStrike's blog adds an important qualifier about the limits of that bypass: in the observed activity, the threat actor periodically had to intervene to overcome the model. Secondary coverage of the method reports that the four groups bypassed safeguards by convincing models that their activities were legitimate, and that the attackers used both open- and closed-weight models, in almost all cases older versions rather than frontier models. Palo Alto Networks and PHP2Shell. The vendor advisory itself lists the product status: PAN-OS versions 12.2, 12.1, 11.2, 11.1 and 10.2 are all listed as affected, with unaffected fix versions listed as "None (ETA: TBD)" for each. The advisory says the issue will be fixed in upcoming releases of PAN-OS and will be updated with fixed versions and release estimates as they become available. Context the post omits sits in the same advisory: Palo Alto rates the CVSS score of this issue in its product as HIGH at 8.7 with suggested urgency MODERATE, and states it is not aware of any malicious exploitation of this issue against its products, plus advance virtual patching coverage for some customers via Threat ID 543004 and a mitigation of restricting management interface access to trusted internal IP addresses. Trade coverage adds that seven Palo Alto products are unaffected: Bridgecrew, Cloud NGFW, Cortex XDR Agent, Cortex XSOAR, Cortex Xpanse, Idira and Prisma Access. Banks. The Axios report is the single originating source. Axios reports that Morgan Stanley has begun blocking third-party agents from accessing its platforms according to sources familiar with the decision, and that JPMorgan, Bank of America and Citi are also restricting agents. Reuters' summary of that report gives the list the post uses: banks including Morgan Stanley, Bank of America, JPMorgan, Wells Fargo and others are racing to contain agent risks, Axios reported on Thursday, citing people familiar with the matter. The on-record bank statements run in the other direction from the alarm: a Chase spokesperson said customers should not give their usernames and passwords to anyone, and a Morgan Stanley spokesperson said the bank hasn't observed instances of "rogue AI agents accessing our platforms."