🔍 Read the full analysis: Cybersecurity In A World Where AI Speeds Up Attacks on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
U.S. agencies and Microsoft have reported cases in which attackers used AI-generated scripts or an agentic framework in operations targeting industrial controllers and cloud services. The reports document faster attempts, but Microsoft says most tenant targeting failed and that a successful sign-in does not prove data was accessed. OpenAI separately forecasts a sharp rise in AI-found vulnerabilities, adding pressure on defenders to patch and secure basic systems.
A joint advisory published August 18 by the NSA, CISA, FBI, Department of Energy and Environmental Protection Agency says threat actors used AI-generated scripts against internet-reachable Siemens S7 programmable logic controllers, or PLCs. These industrial computers can control equipment such as pumps, valves and production lines. The agencies say the activity involved reconnaissance and capability development across sectors including manufacturing, energy, water and food and agriculture.
Microsoft Threat Intelligence reported October 21 that the China-state-aligned actor it calls Storm-0940 used ARTEX, an open-source multi-agent penetration-testing framework, in activity connected to a ConnectWise Cloud Backup compromise. Microsoft says the actor attempted logins against more than 2,250 tenants using credentials associated with the cloud backup application. In a password-spraying campaign from October 10 to 14, Microsoft recorded more than 151,000 failed logins, with a peak of 2,900 per minute. None of those attempts succeeded.
Microsoft says that in several attacks, the time from initial targeting to privileged cloud access was under three hours. It also cautioned that most targeting was unsuccessful and that an affected application or successful sign-in does not by itself show that data was accessed. ConnectWise said existing backup archives showed no indication of compromise and that it suspended Cloud Backup, SaaS Security and Cloud Management while investigating.
Infographic 01 · The record so far
AI is in the attacker’s toolchain. On the record.
Six documented events in twelve weeks. Each label says what the source itself states about AI. Figures are the vendors’ and agencies’ own.
- 30 Jul 2026Unit 42: AI on attackDeepSeek inside the Hermes Agent framework picks and switches targets on its own. Limited impact: across autonomous and manual attempts, data theft from 3 NetScaler systems confirmed. Single-vendor report.
- 18 Aug 2026NSA, CISA, FBI, DOE, EPA: AI on attackAI-generated exploitation scripts, disguised as monitoring tools, against Siemens S7 PLCs in US critical infrastructure. “Not a theoretical risk.”
- 10–14 OctStorm-0940 password spray: AI on attack151,000+ failed logons across eight organizations. No successful sign-in. Lockouts made up nearly 98% of responses.
- 16–21 OctConnectWise Cloud Backup: AI on attackStolen app credentials, then machine-speed sign-ins to customer tenants. Microsoft attributes it to the China-state aligned Storm-0940.
- 21 Oct 2026OpenAI, Patch the Planet: AI on defenceAI found the critical PHP flaws before attackers did. OpenAI expects similar findings to “increase sharply in the coming months.”
- 21–22 OctQuest Diagnostics: no AI link reportedUnauthorized access to a lab ordering and results system; certain patient records accessed. No cause or attacker named.
Most AI-driven attempts in these reports failed. What changed is the cost of trying.Assessment: Thorsten Meyer
Sources: Unit 42 (30 Jul 2026); NSA, CISA, FBI, DOE, EPA advisory (18 Aug 2026); Microsoft Threat Intelligence (21 Oct 2026); ConnectWise Trust Center (from 16 Oct 2026); OpenAI, Patch the Planet (21 Oct 2026); Quest Diagnostics Form 8-K (22 Oct 2026).
Powered by Thorsten Meyer AI
Automation Compresses Attack Timelines
The reports point to a change in the cost and pace of routine attack work, rather than evidence of a new kind of all-powerful AI attacker. Scanning, password spraying, adapting public exploits and testing access are established techniques. Software agents can perform such tasks continuously, potentially allowing an operator to cover more targets or move faster than a person working manually.
That speed can narrow the time defenders have to detect suspicious activity and contain an intrusion. It can also increase the volume of alerts and failed login attempts security teams must review. But the figures do not mean attacks are generally succeeding: Microsoft’s reported password-spray attempts all failed, and it says most targeting in the campaign was unsuccessful.
AI may also increase defensive workload by helping researchers find flaws. OpenAI said AI had identified critical PHP vulnerabilities that were patched on October 21, before attackers found them, and forecast that similar discoveries would “increase sharply in the coming months.” That is a forecast, not a measured estimate of how many flaws will emerge or how quickly organizations can patch them.
hardware security key FIDO2 USB-C phishing-resistant MFA
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Three Reports, Different Evidence
The August advisory describes activity against Siemens S7 PLCs. According to the agencies, attackers used internet scanning services to locate exposed controllers and AI-generated exploitation scripts disguised as monitoring tools, built with the open-source snap7 library. The agencies assess that the activity may be preparation to cause operational effects; that assessment does not mean operational disruption has been confirmed.
Microsoft’s October report concerns a separate software supply-chain route: use of access associated with ConnectWise Cloud Backup to attempt entry into customer tenants. Microsoft attributes the activity to Storm-0940 and says ARTEX can perform tasks from reconnaissance through data exfiltration with limited human interaction. Those descriptions are Microsoft’s assessment of the actor and framework, not proof that data exfiltration occurred in every targeted environment.
A July 30 report from Palo Alto Networks’ Unit 42 described another case involving an actor using the DeepSeek model with the Hermes Agent framework and directing it through Telegram. Unit 42 said the agent searched GitHub after an initial exploit attempt failed. Separately, a security leader cited in the source material predicted worsening conditions and a “vulnerability tsunami” within months. That forecast is an individual expert opinion, not a measured finding or an agency projection.
““This is not a theoretical risk—it is an active threat.””
— Joint cybersecurity advisory by the NSA, CISA, FBI, DOE and EPA
small business firewall with intrusion prevention system
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Impact and Scale Still Unclear
The public reports do not establish how many industrial controllers were accessed, whether the activity led to operational effects, or whether all affected organizations have been identified. The agencies describe the PLC activity as active and assess it as likely preparation for possible effects, but they do not report confirmed disruption in the supplied material.
For the ConnectWise-related campaign, Microsoft says a successful sign-in does not necessarily mean data was accessed. ConnectWise reported no indication of compromise in existing backup archives, but the investigation was ongoing when the source material was published. The material also does not provide a verified count of organizations that suffered data exposure or explain the full scope of any access.
More broadly, the cited cases do not establish how often AI is being used in attacks compared with conventional tools, or how much automation contributed to each operation’s success. Predictions of a coming surge in vulnerabilities remain forecasts, not confirmed outcomes.
managed network switch VLAN support for IoT and industrial devices
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Investigations and Basic Controls
ConnectWise said it had suspended the named services while it investigated. The source material does not give a date for their return or a final investigation report. Organizations using those services should follow updates from ConnectWise and review relevant account activity and access permissions.
For industrial operators, the joint advisory recommends inventorying controllers, patching firmware, keeping PLCs off the public internet, blocking TCP port 102 at the perimeter and separating IT and operational technology networks. It also calls for multi-factor authentication on remote OT access and monitoring S7 traffic for writes outside approved change windows.
For companies more broadly, the practical next steps are to maintain a current asset inventory, set patch deadlines, use phishing-resistant sign-in where available, review third-party application access and practise incident response. Individuals can begin by securing their email account, which often provides access to password resets and other services. Further reporting should clarify the ConnectWise investigation’s findings and whether agencies identify additional affected infrastructure.
industrial Ethernet firewall for PLC network segmentation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did U.S. agencies report about AI and industrial systems?
A joint advisory says actors used AI-generated scripts against internet-reachable Siemens S7 programmable logic controllers. The agencies describe reconnaissance and capability development and assess the activity as likely preparation for possible operational effects; the supplied material does not confirm disruption.
Did the Microsoft-reported campaign compromise more than 2,250 organizations?
No. Microsoft said the actor attempted logins against more than 2,250 tenants. It reported that all password-spray attempts in the cited campaign failed and said most targeting was unsuccessful. A targeted tenant is not necessarily a compromised tenant.
Was data stolen through the ConnectWise activity?
The supplied reports do not confirm data theft. Microsoft said a successful sign-in or an affected app does not necessarily mean data was accessed. ConnectWise said existing backup archives showed no indication of compromise while its investigation continued.
What should organizations do in response?
The advisory and reports support basic controls: keep industrial controllers off the public internet, patch firmware, separate IT and OT networks, require strong multi-factor authentication for remote access, review third-party app permissions and monitor account activity. Organizations should also maintain asset inventories and practise incident response.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
