🔍 Read the full analysis: AI And Cyberattacks: How Companies And Users Can Adapt on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
US agencies say attackers have used AI-generated scripts against internet-reachable industrial controllers, while Microsoft reports an AI agent was used in attacks involving a compromised cloud backup app. Most attempts in Microsoft’s account failed, and the companies and agencies have not established that all targeted organizations suffered data access. The reports point to familiar defenses: secure exposed systems, tighten account and app access, patch promptly, and rehearse incident response.
US agencies and Microsoft have reported separate cases of AI-assisted cyber activity targeting industrial control systems and cloud accounts, showing how attackers can use software to speed up familiar steps such as reconnaissance, exploitation attempts and password spraying. The reports do not show that every target was breached: Microsoft says most of the activity it tracked was unsuccessful, while the agencies warn that some industrial systems remain exposed to attempts.
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. PLCs are industrial computers that control equipment and processes. The advisory describes scripts built with the open-source snap7 library and disguised as monitoring tools. It names S7 models from the S7-200 through S7-1500, including safety controllers, and says the activity could reduce the expertise and time needed to attempt exploitation.
Microsoft Threat Intelligence reported October 21 that it linked the China-state-aligned actor Storm-0940 to use of ARTEX, an open-source multi-agent penetration-testing framework, in activity involving ConnectWise Cloud Backup credentials. Microsoft says the actor attempted sign-ins to more than 2,250 tenants. It also recorded more than 151,000 failed logons during a password-spraying campaign from October 10 to 14, with a peak of 2,900 failures per minute; Microsoft said none of those attempts succeeded.
Microsoft said several attacks reached privileged cloud access in under three hours, but cautioned that most targeting attempts failed and that a successful sign-in or affected app does not by itself establish that data was accessed. ConnectWise disclosed the compromise on October 16 and said existing backup archives showed no indication of compromise. The company 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
Faster Attacks Raise the Cost of Delay
The reports matter because they describe automation applied to routine attack tasks, not a new class of unstoppable software. A program can scan systems or try credentials repeatedly without fatigue, potentially giving defenders less time to notice activity and respond. In Microsoft’s account, some cases moved from targeting to privileged access in less than three hours. That is a report about specific incidents, not proof that all AI-assisted attacks move at that speed.
Industrial systems carry a distinct risk: access to a controller can affect physical operations, not just files or accounts. The agencies say the activity may be preparation for operational effects, an assessment rather than confirmation that disruption occurred. For cloud customers, the ConnectWise episode shows how a trusted third-party application’s access can extend an incident across tenants. These risks make asset inventories, limited privileges and monitoring of vendor connections relevant even when an organization does not build or deploy AI.
AI can also help defenders find flaws. OpenAI said AI identified critical PHP vulnerabilities that were patched on October 21, before attackers exploited them, and forecast that AI-discovered vulnerabilities will increase sharply in coming months. That is a company forecast, not a measured estimate of future attack volume. More reported flaws could also mean greater pressure on organizations to prioritize and install fixes.
hardware security key FIDO2 USB-C multi-factor authentication
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Two Documented Paths Into Systems
The industrial advisory and Microsoft report describe different targets and methods. In the first, agencies say actors searched for exposed Siemens controllers and used AI-generated scripts based on snap7. Their recommendations include inventorying PLCs, updating firmware, keeping controllers off the public internet, blocking TCP port 102 at network boundaries, and separating operational technology from corporate IT. They also recommend multi-factor authentication for remote OT access and monitoring for S7 write commands outside approved change windows.
In Microsoft’s account, the route involved credentials for a cloud backup application and trust already granted by customers. The company says ARTEX can coordinate activity from reconnaissance through data exfiltration with limited human involvement. That description does not establish that the framework completed every stage in each incident. ConnectWise’s reported suspension of affected services and its statement about backup archives are company disclosures; the investigation’s final findings were not included in the source material.
A separate July 30 report from Palo Alto Networks’ Unit 42 described a Chinese-speaking actor using the DeepSeek model with the Hermes Agent framework, directed through Telegram. Unit 42 said the agent searched GitHub after its first exploit attempt failed. The available source material does not provide the report’s full findings, so this case cannot support broader conclusions about how often such agents succeed.
“The majority of the targeting wasn’t successful.”
— Microsoft Threat Intelligence
Access Does Not Prove Data Theft
The extent of any data access remains unclear in the ConnectWise-related activity described by Microsoft. Microsoft says a successful sign-in or an affected application does not necessarily mean data was accessed, and ConnectWise said existing backup archives showed no indication of compromise. The source material does not include a final investigation report or a full account of which tenants were affected.
For the industrial activity, the agencies assess that actors are likely preparing to cause operational effects, but the material does not confirm a resulting disruption to infrastructure. The number of controllers successfully accessed, the identities of affected operators and any resulting physical effects are not specified. Separately, a security leader’s prediction of a near-term “vulnerability tsunami” and an roughly 18-month period to prepare is an expert opinion, not a measured forecast or official finding.
Investigations and Routine Defenses
ConnectWise’s services were suspended while the company investigated; the source material does not give a restoration date or final findings. Organizations using the affected services should follow vendor updates and review their own access and activity records rather than assume that an attempted sign-in proves data exposure.
For companies, the immediate work is concrete: maintain an inventory of internet-facing systems and industrial controllers, set patch deadlines, require phishing-resistant sign-in where available, review third-party applications and their permissions, and separate OT networks from business IT. Teams should also test incident-response plans and monitor remote access and unusual controller commands. Users can begin with their email account: use a unique password, enable multi-factor authentication, check recovery details and review connected apps and active sessions.
The agency and company reports document current activity, while the wider scale and future pace of AI-assisted attacks remain uncertain. The next useful evidence will be investigation findings, confirmed disclosures from affected organizations and further advisories that identify whether the same methods are spreading.
Key Questions
Do the reports show that AI cyberattacks succeeded against every target?
No. Microsoft said most targeting was unsuccessful and reported that all more than 151,000 password-spray attempts in the campaign it described failed. It also said successful sign-in does not necessarily mean data was accessed.
What systems did the US agencies say were targeted?
The joint advisory describes attempts against internet-reachable Siemens S7 programmable logic controllers, which are used to control industrial equipment and processes. It names models from the S7-200 to the S7-1500.
What should a company do first?
Start by identifying exposed systems and accounts, then address known vulnerabilities, restrict remote access, require strong multi-factor authentication, and review third-party app permissions. Operators of industrial systems should keep PLCs off the public internet and separate OT networks from corporate IT.
Does an AI-discovered vulnerability mean an attack is underway?
No. Finding a vulnerability is not evidence that attackers exploited it. OpenAI said AI identified critical PHP flaws that were patched on October 21 before attackers did; its forecast of more such findings is not a count of active attacks.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
