How Users Can Respond To Cyberattacks Moving Faster With AI
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: How Users Can Respond To Cyberattacks Moving Faster With AI on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

US agencies and cybersecurity companies have documented attackers using AI-generated scripts and agentic tools for reconnaissance, password spraying and attempts to exploit trusted software access. The reports show faster attack activity, not that most attempts succeed; practical defenses remain account security, patching, access reviews and practiced response plans.

US agencies and cybersecurity researchers have documented attackers using AI-generated scripts and autonomous software agents in attempts to target industrial controllers and cloud services, including activity Microsoft linked to a supply-chain compromise. The reports do not show that most attempts succeeded, but they describe software carrying out familiar attack steps faster and with less direct human effort, raising the pressure on companies and individual users to secure accounts, patch systems and review trusted access.

A joint advisory published Aug. 18 by the NSA, CISA, FBI, Department of Energy and Environmental Protection Agency says actors used AI-generated scripts disguised as monitoring tools against internet-reachable Siemens S7 programmable logic controllers. PLCs are industrial computers used to control equipment and processes. The agencies describe reconnaissance and capability development targeting sectors including manufacturing, energy, water, chemicals and agriculture, and assess that the activity may be preparation for operational effects. They recommend keeping controllers off the public internet, maintaining firmware, separating operational technology from business networks and monitoring access.

In an Oct. 21 report, Microsoft Threat Intelligence linked China-state-aligned actor Storm-0940 to use of ARTEX, an open-source agentic penetration-testing framework, in activity involving the ConnectWise Cloud Backup compromise. Microsoft said the actor tried credentials against more than 2,250 tenants and recorded more than 151,000 failed logons between Oct. 10 and 14, with a peak of 2,900 failures per minute. The company said some attacks reached privileged cloud access in under three hours. It also stressed that most targeting was unsuccessful and a successful sign-in does not by itself establish that data was accessed.

ConnectWise disclosed the compromise on Oct. 16 and said existing backup archives showed no indication of compromise. The company suspended Cloud Backup, SaaS Security and Cloud Management while investigating. Separately, OpenAI said AI had identified critical PHP vulnerabilities that were patched on Oct. 21 before attackers exploited them, and forecast that AI-found critical vulnerabilities will increase sharply in coming months. That is a forecast, not a count of future vulnerabilities.

At a glance
reportWhen: Reports published from July through Oct…
The developmentRecent government advisories and threat reports describe AI-assisted attacks against industrial controllers and cloud environments, prompting renewed calls for basic security measures.
Thorsten Meyer AI
Primary sources · Jul–Oct 2026

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.

<3 hfrom first targeting to privileged cloud access, in several Storm-0940 attacks
2,250+Microsoft 365 tenants the ARTEX agents tried to sign in to
2,900failed logons per minute at the peak of one spray. Successes: 0
h → min“hundreds of hours of manual targeting analysis” done in minutes (Unit 42)
  1. 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.
  2. 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.”
  3. 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.
  4. 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.
  5. 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.”
  6. 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 point to a change in the speed and scale of routine attack work, rather than evidence of an entirely new kind of attacker. Software can help scan targets, test credentials or adapt publicly available techniques, and can keep working without the fatigue or time constraints of a human operator. Microsoft reported rapid movement from initial targeting to privileged access in some cases, while emphasizing that most attempts did not succeed.

For organizations, that combination makes basic safeguards more consequential. A compromised vendor app or exposed controller can provide an entry point that bypasses some users’ expectations about where risk lies. For individuals, email and other identity accounts matter because password reuse, weak sign-in controls or an unreviewed app connection can make one account a route into others. AI does not remove the need for familiar defenses; it can reduce the time available to apply them after suspicious activity begins.

There is also a defensive side. OpenAI’s account of AI-assisted vulnerability discovery suggests that more flaws may be found and fixed before attackers exploit them. But that prospect brings a practical burden: organizations may face more patches to assess and deploy, while the advantage depends on whether affected systems are identified and updated promptly. The source material does not establish how often AI discovery will lead to fixes ahead of exploitation.

Products Worth Considering

Jackery Explorer 1000 v2 Portable Power Station, 1024Wh LiFePO4 Battery

Jackery Explorer 1000 v2 Portable Power Station, 1024Wh LiFePO4 Battery

As an affiliate, we earn on qualifying purchases.

From Industrial Controllers to Cloud Accounts

The activity described in the reports covers different systems and should not be treated as one coordinated campaign. The Aug. 18 government advisory concerns Siemens S7 controllers and operational technology. Microsoft’s Oct. 21 report concerns Storm-0940, ARTEX and access attempts involving the ConnectWise Cloud Backup service. The common thread is the use of AI-related tools to support known steps such as reconnaissance, credential testing and exploitation.

Microsoft said Storm-0940 has been active since 2021 and assessed the actor was behind the ConnectWise-related activity. Its figures are specific to the reported campaign: more than 151,000 failed logons from Oct. 10 to 14, not a general measure of the rise in AI-driven attacks. Microsoft also said ConnectWise app credentials were used to try access across tenants that had trusted the app. Affected application access does not automatically mean customer data was obtained.

Another example cited in the source material is a July 30 report from Palo Alto Networks’ Unit 42 describing an actor using the DeepSeek model through the Hermes Agent framework and Telegram. The provided report excerpt says that after an initial exploit failed, the agent searched GitHub. It does not include enough detail to establish the full outcome of that incident. Such examples illustrate reported experiments and activity, but do not by themselves quantify how widespread or effective agent-led attacks are.

“This is not a theoretical risk—it is an active threat.”

— Joint Cybersecurity Advisory from the NSA, CISA, FBI, Department of Energy and EPA, published Aug. 18

Success Rates and Future Scale

The reports do not establish the overall prevalence of AI-assisted attacks, or show that AI was necessary for the activity described. Microsoft’s account says most targeting failed; ConnectWise said its existing backup archives showed no indication of compromise. The available material does not specify whether customer data was accessed in the reported cases, or provide a complete account of the investigation’s findings.

The government advisory describes activity against controllers and assesses possible preparation for operational effects, but the supplied material does not confirm that physical disruption occurred. OpenAI’s prediction of a sharp rise in AI-found critical flaws remains a forward-looking claim. A separate forecast that the threat may worsen within months and that organizations have roughly 18 months to prepare is an individual security leader’s opinion, not a measured timeline or official assessment.

Patch, Review Access, Practice Response

Organizations should start with systems and access they already rely on: keep an accurate inventory, set and meet patch deadlines, require phishing-resistant sign-in where available, and review third-party applications and their permissions. For industrial operators, the agencies’ recommendations include keeping PLCs unreachable from the internet, separating operational technology from corporate networks, applying firmware updates and monitoring controller traffic for unexpected changes.

Companies should also rehearse how they will contain a compromised account, vendor app or cloud service, and check that incident-response plans work across security, IT and operational teams. Users can begin with their email account: use a unique password, turn on multifactor authentication, review recovery details and remove app connections they no longer need. These steps do not guarantee protection, but they reduce common routes into accounts and services.

For ConnectWise customers, the next useful information is the company’s investigation findings and any specific guidance about affected services, accounts or data. Microsoft’s campaign reporting, the agencies’ industrial guidance and any future vulnerability disclosures may add detail. Until then, the confirmed reports support a practical response: treat AI as a speed multiplier for familiar threats, and make sure basic controls are in place before an incident.

Key Questions

Do the reports show that AI attacks are usually successful?

No. Microsoft said most of the targeting it described was unsuccessful. Its report also cautions that an affected app or successful sign-in does not by itself prove data access.

What systems did the government advisory describe?

The Aug. 18 advisory describes AI-generated scripts used against internet-reachable Siemens S7 programmable logic controllers, which control industrial equipment and processes.

What should individual users do first?

Secure your email account with a unique password and multifactor authentication. Review its recovery information and connected apps, and remove access you no longer use.

Is a coming increase in AI-found vulnerabilities confirmed?

No. OpenAI said it expects AI-found critical vulnerabilities to increase sharply in coming months. That is a forecast, not an observed future result or a quantified estimate in the material provided.

Has ConnectWise confirmed that customer data was stolen?

The supplied material does not confirm customer data theft. ConnectWise said existing backup archives showed no indication of compromise and suspended several services while it investigated.

Source: ThorstenMeyerAI.com

You May Also Like

Generative AI in Phishing and Scams: Emerging Threats and Solutions

Understanding how generative AI fuels evolving phishing scams reveals why staying vigilant is more critical than ever.

Anthropic Claude Extensions: What The Supply Chain Attack Involved

A WIRED headline reports malicious Claude extensions reached thousands of devices, but the available material does not verify the count, method or impact.

The August 1 Strategy: Using AI Benchmarks To Strengthen National Security Secrets

The US government will implement classified AI capability benchmarks by August 1, affecting developers and national security, with voluntary pre-release assessments.

A Thousand Fewer Paths To Root

OpenAI says a prototype found 2,308 Linux kernel candidates, reproduced 1,724 vulnerabilities and generated 948 local privilege escalation exploits.