🔍 Read the full analysis: A Closer Look At AI’s Role In Faster Cyberattacks on ThorstenMeyerAI.com
Get the latest gadgets 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
US agencies and cybersecurity researchers have documented attackers using AI-generated scripts and agent frameworks in operations involving industrial controllers and cloud services. Microsoft reported rapid targeting and some access to privileged cloud accounts, but said most targeting attempts failed; the scope of any data access remains unclear.
US agencies and cybersecurity firms have reported attackers using AI-generated scripts and agent frameworks in operations against industrial controllers and cloud services, documenting a shift toward automating familiar attack steps. Microsoft said one campaign reached privileged cloud access in several cases in under three hours, while also reporting that most of its observed targeting attempts failed.
A joint advisory published August 18 by the NSA, CISA, FBI, Department of Energy and Environmental Protection Agency describes activity involving Siemens S7 programmable logic controllers, or PLCs, used to operate industrial processes. The agencies said actors scanned for internet-reachable controllers and used AI-generated exploitation scripts disguised as monitoring tools. The advisory named models from the S7-200 through the S7-1500, including safety controllers, and assessed that the activity was likely preparation for operational effects against critical infrastructure.
On October 21, Microsoft Threat Intelligence reported that Storm-0940, which Microsoft describes as a China-state-aligned actor active since 2021, used the open-source ARTEX multi-agent framework in activity connected to a ConnectWise Cloud Backup compromise. Microsoft said the actor used credentials for the backup application to attempt sign-ins to more than 2,250 customer tenants. Between October 10 and 14, Microsoft recorded more than 151,000 failed logons, peaking at 2,900 per minute. The company said the majority of targeting was unsuccessful. In several attacks, it said, the period from initial targeting to privileged cloud access was under three hours.
Microsoft cautioned that an affected application or successful sign-in does not by itself establish that data was accessed. ConnectWise said its existing backup archives showed no indication of compromise and said it suspended Cloud Backup, SaaS Security and Cloud Management while investigating. Separately, Palo Alto Networks’ Unit 42 described an AI agent performing hundreds of hours of manual targeting analysis in minutes. These reports document particular cases, not a measure of how common AI-assisted attacks are across all cybercrime.
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 speed and scale as the immediate security concern, rather than evidence of a wholly new kind of attacker. Scanning for exposed systems, trying credentials, exploiting vulnerabilities and using trusted applications are established tactics. AI tools may allow operators to run more of those tasks, for longer periods, or with less manual effort.
That can shorten the time defenders have to detect unusual activity and respond. An attack that moves from initial targeting to privileged access in less than three hours leaves a narrow window for alert review, account containment and investigation. For operators of industrial systems, a compromised controller can also pose risks beyond data loss, although the advisory describes likely preparation for operational effects rather than confirming that physical disruption occurred.
Automation can also increase pressure on defenders from the other direction. OpenAI said AI helped identify critical PHP vulnerabilities that were patched on October 21 and forecast that AI-discovered critical vulnerabilities will “increase sharply in the coming months.” That is a forecast, not a count of future flaws. If more vulnerabilities are found, organizations may face more urgent decisions about testing and applying patches, alongside the need to investigate active attacks.
Products Worth Considering

LG 27GR83Q-B 27-inch Ultragear QHD IPS Gaming Monitor

Samsung 27-Inch Odyssey G5 QHD Gaming Monitor (G50F)

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)
As an affiliate, we earn on qualifying purchases.
Three Reports, Different Evidence
The source material describes several separate developments, rather than one coordinated incident. The industrial-control advisory is a joint government warning about activity against Siemens PLCs. Microsoft’s October report concerns Storm-0940, ARTEX and activity tied to a third-party cloud backup service. Unit 42’s July 30 report describes a separate actor using a language model through the Hermes Agent framework and Telegram. These cases should not be treated as proof that the same group or tool was involved in every operation.
The accounts also differ in what they establish. Government agencies describe reconnaissance and capability development and assess that actors may be preparing for operational effects. Microsoft reports attempted logons, some instances of rapid privileged access and a compromise it links to Storm-0940, while stressing that most targeting failed and that access does not necessarily mean data was obtained. Observed activity, an agency assessment and a forecast are distinct kinds of evidence.
Microsoft described ARTEX as an open-source framework capable of supporting stages from reconnaissance to data exfiltration with limited human interaction. That description concerns the framework’s capabilities; it does not mean every operation completed every stage. The source material also cites an expert’s expectation of a coming “vulnerability tsunami” and an approximately 18-month period of strain. Those are opinions about what may happen, not measured trends or confirmed timelines.
““This is not a theoretical risk—it is an active threat.””
— Joint advisory by the NSA, CISA, FBI, Department of Energy and EPA
The Extent of Access Remains Unclear
Public reporting does not establish how much data, if any, was taken from cloud tenants in the activity Microsoft described. Microsoft said a successful sign-in does not necessarily indicate data access; ConnectWise said its existing backup archives showed no indication of compromise, but suspended services during its investigation. The available information does not settle the full impact on every affected customer or service.
The reports also do not quantify the prevalence of AI-assisted operations compared with attacks that rely on conventional automation. They document specific cases and assessments, but do not show that AI is responsible for a general increase in attack success rates. Nor do they confirm that the PLC activity caused physical disruption. Predictions of sharply rising vulnerability discoveries or a “vulnerability tsunami” remain forecasts, not established outcomes.
Investigations and Basic Defenses Continue
ConnectWise’s investigation will help clarify the effect of the cloud service compromise and whether customer data was accessed. Organizations using the affected services will need to review vendor updates and any guidance on account security or possible exposure. The government advisory directs industrial operators to inventory controllers, apply firmware updates, prevent direct internet access to PLCs, separate IT and operational technology networks, require multifactor authentication for remote access and monitor controller traffic for unexpected writes.
For companies more broadly, the source material recommends focusing on established safeguards rather than buying AI tools as a substitute for security fundamentals: maintain an accurate asset inventory, set patch deadlines, use phishing-resistant sign-in where possible, review third-party application access and practise incident response. For individuals, securing the email account is a practical starting point because it can be used to reset access to other services. Further incident details and vulnerability disclosures may change the risk picture; the reports do not provide a date for when all investigations will be complete.
Key Questions
A joint advisory from five US agencies says actors used AI-generated scripts in activity targeting internet-reachable Siemens S7 PLCs. The agencies described reconnaissance and capability development and assessed that the activity was likely preparation for operational effects; they did not report confirmed physical disruption in the material provided.
Did the Storm-0940 activity compromise more than 2,250 tenants?
Microsoft said the actor attempted sign-ins to more than 2,250 tenants using credentials associated with the cloud backup application. That figure describes attempted targeting, not confirmed compromises. Microsoft said most targeting was unsuccessful and that a successful sign-in does not necessarily mean data was accessed.
Does this show AI attacks are more successful?
The reports show AI being used in particular operations and describe how it can support faster or more extensive work. They do not establish a general increase in attack success rates or quantify how common AI-assisted attacks are across the wider threat landscape.
What should organizations do now?
Follow established security practices: inventory systems, patch software and firmware, restrict internet access to industrial controllers, separate IT and operational technology networks, review third-party app permissions, require strong authentication and practise incident response. These measures address risks described in the reports without depending on predictions about future AI capabilities.
Source: ThorstenMeyerAI.com
