AI-Powered Attacks Move Quickly: What Should You Do?

🔍 Read the full analysis: AI-Powered Attacks Move Quickly: What Should You Do? on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get smart everyday buys delivered free with Prime

• Fast, free delivery on millions of items
• Prime Video, Amazon Music and more included
• Member-only deals all year

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

U.S. agencies and Microsoft have reported separate cases of AI-assisted cyber activity involving industrial control systems and cloud accounts, showing how automated tools can speed up familiar attack methods. The agencies say actors used AI-generated scripts against Siemens controllers, while Microsoft says a China-state-aligned group used an open-source agent framework in attacks involving a compromised cloud-backup application. Most of the targeting Microsoft described did not succeed, and the reports do not establish that every affected account suffered data theft.

A joint August 18 advisory from the NSA, CISA, FBI, Department of Energy and Environmental Protection Agency says threat actors have used AI-generated exploitation scripts against internet-reachable Siemens S7 programmable logic controllers. PLCs are industrial computers used to control equipment and processes. The advisory says actors identified exposed devices through internet-scanning services and disguised scripts built with the open-source snap7 library as legitimate monitoring tools. The agencies assess that the activity may be preparation for operational effects against critical infrastructure; that is an assessment, not confirmation that disruption occurred.

In an October 21 report, Microsoft Threat Intelligence said the actor Storm-0940 used ARTEX, an open-source multi-agent penetration-testing framework, in activity involving ConnectWise Cloud Backup. Microsoft says the actor attempted logins across more than 2,250 tenants after obtaining credentials for the application. It also recorded more than 151,000 failed logons during a password-spray campaign from October 10 to 14, with a peak of 2,900 failures per minute; Microsoft said none of those attempts succeeded. In several attacks, the time from initial targeting to privileged cloud access was under three hours.

Microsoft cautioned that most targeting was unsuccessful and that an affected application or successful sign-in does not necessarily mean data was accessed. ConnectWise, which disclosed the compromise on October 16, said existing backup archives showed no indication of compromise and suspended Cloud Backup, SaaS Security and Cloud Management while investigating. Separately, OpenAI said AI had identified critical PHP vulnerabilities that were patched on October 21, and forecast that AI-found vulnerabilities would increase sharply in coming months. That is a forecast, not a measured prediction of how many flaws will be found or exploited.

Automation Compresses Attack Timelines

The reports matter because they describe automation applied to established attack steps, not evidence of a new class of all-powerful AI attacker. Scanning for exposed equipment, attempting passwords and exploiting trusted software are known techniques. AI agents and generated scripts can help perform or adapt those tasks at scale and outside normal working hours, leaving defenders less time to detect activity and respond.

For industrial operators, an internet-accessible controller can create a route toward systems that affect physical processes. For cloud customers, a trusted application can carry permissions into many organizations, so a compromise of the supplier may have a wider reach than a single account incident. The consequences differ by target, and the reports do not say that every attempt caused damage. They do underline the value of limiting exposure and permissions before an incident.

AI can also assist defenders. OpenAI’s account of AI-supported vulnerability discovery points to the possibility of more flaws being found and fixed. That may increase the pace of patching work, but it does not show that every organization will receive or apply fixes promptly. The practical challenge is to maintain an accurate asset inventory, prioritize exposed and high-impact systems, and have a response process ready when alerts arrive.

Three Reports, Different Evidence

The source material describes three strands of reporting, with distinct levels of detail. The government advisory names Siemens S7 PLCs, including models from the S7-200 through S7-1500, and says the activity affects sectors such as manufacturing, energy, water, chemicals, food and agriculture, and commercial facilities. Its recommendations include inventorying controllers, applying firmware updates, preventing direct internet access, blocking TCP port 102 at the perimeter, separating IT and operational technology networks, using multi-factor authentication for remote OT access and monitoring S7 traffic for writes outside approved change windows.

Microsoft’s October report concerns Storm-0940, which it describes as a China-state-aligned actor active since 2021, and the ConnectWise Cloud Backup compromise. Microsoft says ARTEX can support activity from reconnaissance through data exfiltration with limited human interaction. The reported tenant attempts and rapid access in some cases describe activity Microsoft observed; they do not establish that every targeted tenant was breached or that data was taken.

A July 30 Unit 42 report, as summarized in the source material, describes a Chinese-speaking actor using the DeepSeek model with the Hermes Agent framework and steering it through Telegram. The supplied account says the agent searched GitHub after its first exploit failed, but the source text ends before explaining what it found or the outcome. That detail cannot support a conclusion about the operation’s success.

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

— Joint advisory by the NSA, CISA, FBI, Department of Energy and EPA

Breach Scope Remains Unsettled

It is not clear from the reports how many organizations experienced confirmed data access in the ConnectWise-related activity. Microsoft says a successful sign-in does not necessarily mean data was accessed, while ConnectWise said its existing backup archives showed no indication of compromise. The investigation status and any later findings are not provided in the supplied material.

The government advisory describes activity against exposed controllers and assesses possible preparation for operational effects. It does not establish that the actors caused physical disruption. Nor do the reports quantify how much AI contributed compared with other tools or human decisions in each operation. OpenAI’s expected rise in AI-identified vulnerabilities is a forecast, and a separate security leader’s prediction of a near-term “vulnerability tsunami” and roughly 18-month period of pressure is an attributed opinion, not a measured outlook.

Defensive Steps and Investigations

Organizations should act on the specific steps in the advisory and Microsoft report rather than wait for evidence of a wider trend. For industrial operators, that means locating every PLC, patching firmware, removing direct internet exposure, restricting remote access with multi-factor authentication and separating operational technology from corporate networks. The agencies also recommend blocking TCP port 102 at the perimeter and monitoring for unauthorized S7 writes.

Cloud customers should review third-party applications and the permissions they hold, remove access that is not needed, and confirm how suppliers will report and investigate incidents. Security teams should prioritize internet-facing and high-impact systems for patching, use phishing-resistant sign-in where available, and practise incident response. Individual users can begin by securing their email account: use a unique password, enable multi-factor authentication and review account recovery and forwarding settings.

Further updates depend on investigation findings from ConnectWise and any additional advisories from government agencies or threat researchers. The source material gives no date for a final ConnectWise report or for a new government assessment. Until more information is available, the confirmed cases support basic exposure reduction and account controls, not predictions about a fixed timetable for future attacks.

Key Questions

Did the reported attacks prove that AI can independently carry out a cyberattack?

No. The reports describe AI-generated scripts and an agent framework being used in attack activity. They do not establish that AI operated independently or without human direction in every case.

Did the ConnectWise-related activity affect all 2,250 tenants?

No. Microsoft said the actor attempted logins across more than 2,250 tenants. It also said most targeting was unsuccessful; the number of attempted tenants is not a count of confirmed breaches.

Was data taken from ConnectWise customers?

The supplied reports do not confirm data theft across affected customers. Microsoft said a successful sign-in does not necessarily mean data was accessed, and ConnectWise said existing backup archives showed no indication of compromise while it investigated.

What should industrial operators do about the PLC warning?

The joint advisory recommends inventorying controllers, patching firmware, removing direct internet access, blocking TCP port 102 at the perimeter, separating IT and OT networks, requiring multi-factor authentication for remote OT access and monitoring S7 traffic for unauthorized writes.

Is a sharp rise in AI-found vulnerabilities confirmed?

No. OpenAI said it expects AI-found critical vulnerabilities to increase sharply in coming months. That is a forecast; the supplied material does not provide a measured future count.

Source: ThorstenMeyerAI.com