The Clock Is Ticking Faster On Cyberattacks In The Age Of AI

🔍 Read the full analysis: The Clock Is Ticking Faster On Cyberattacks In The Age Of AI 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

US agencies and cybersecurity companies have reported attackers using AI-generated scripts and agentic software in campaigns targeting industrial controllers and cloud services, adding evidence that AI is already part of some cyber operations. The reports describe familiar methods carried out at machine speed—not proof of a general-purpose “AI super-hacker”—and include important limits: Microsoft said most of the targeting it observed failed.

A joint advisory issued August 18 by the NSA, CISA, FBI, Department of Energy and Environmental Protection Agency says actors used AI-generated exploitation scripts against internet-accessible Siemens S7 programmable logic controllers, or PLCs. PLCs are industrial computers used to control equipment such as pumps and production lines. The agencies said the activity involved reconnaissance and capability development and could be preparation for operational effects, an assessment rather than confirmation that those effects occurred.

Microsoft Threat Intelligence reported October 21 that Storm-0940, which Microsoft describes as a China-state-aligned actor active since 2021, used the open-source ARTEX multi-agent penetration-testing framework in activity linked to a ConnectWise Cloud Backup compromise. Microsoft said the actor used credentials associated with the cloud backup application to attempt logins across more than 2,250 tenants. From October 10 to 14, Microsoft recorded more than 151,000 failed logons, with a peak of 2,900 per minute; none of those attempts succeeded. In several other attacks described by Microsoft, 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, by itself, show that data was accessed. ConnectWise disclosed the compromise on October 16 and said existing backup archives showed no indication of compromise. It suspended Cloud Backup, SaaS Security and Cloud Management while investigating.

Machine Speed Changes the Defensive Clock

The reports point to a practical change in the pace and cost of cyber operations. AI tools can automate steps such as reconnaissance, password spraying and adapting scripts, allowing attackers to repeat attempts quickly and with less hands-on work. That does not mean every attempt succeeds or that AI independently makes strategic decisions. In Microsoft’s account, a large volume of login attempts failed, even as other activity reportedly reached privileged cloud access in under three hours.

For defenders, the consequence is less time to spot unusual activity and contain a breach. A trusted third-party application can also create a route into many customer environments if its credentials or access are abused. On the other side, AI can help identify software flaws: OpenAI said AI found critical vulnerabilities in PHP that were patched October 21, before attackers did. The company forecast that such AI-discovered flaws would “increase sharply in the coming months.” That is a prediction, not a measured estimate of future incidents, and could bring more pressure to assess and apply patches promptly.

The near-term response described in the source material is largely established security practice, not a request to buy AI tools. Organizations need accurate asset inventories, clear patch deadlines, phishing-resistant sign-in, reviews of third-party applications, separation between corporate IT and operational technology, and rehearsed incident response. For individuals, securing the email account that can reset access to other services is a useful starting point.

Three Reports, Different Evidence

The August government advisory concerns Siemens S7 PLCs, including models from the S7-200 to the S7-1500. The agencies said actors found exposed devices through internet-scanning services and used scripts built with the open-source snap7 library, disguised as legitimate monitoring tools. They identified manufacturing, energy, water, chemical, food and agriculture, and commercial facilities as sectors of concern. Their recommendations include inventorying controllers, updating firmware, removing direct internet exposure, blocking TCP port 102 at network perimeters, separating IT and OT networks, requiring multifactor authentication for remote OT access, and monitoring for unauthorized S7 writes.

The Microsoft report concerns a different path: an AI-enabled framework allegedly used in activity associated with a third-party backup service. Microsoft said ARTEX can support tasks from reconnaissance through data exfiltration with limited human interaction. Its account does not establish that every targeted tenant was accessed or that customer data was taken. ConnectWise’s statement that existing archives showed no indication of compromise is also narrower than a finding that no other systems or data were affected.

Security firm Palo Alto Networks’ Unit 42 described a separate case in a July 30 report, involving an actor who connected the DeepSeek model to the Hermes Agent framework and directed it through Telegram. Unit 42 said the agent performed extensive targeting analysis quickly. The source material also says OpenAI reported AI-assisted identification of critical PHP vulnerabilities patched October 21. These reports show different uses of AI—offensive automation and defensive research—and should not be treated as evidence of one coordinated campaign.

“The majority of the targeting wasn’t successful.”

— Microsoft Threat Intelligence

Access and Impact Remain Unsettled

The reports do not establish how much data, if any, was accessed through the ConnectWise-related activity, or whether any targeted PLCs experienced operational disruption. Microsoft explicitly said an affected application or successful sign-in does not necessarily mean data was accessed. ConnectWise said its existing backup archives showed no indication of compromise, while the investigation was still underway.

The government advisory describes activity the agencies assess as likely preparation for operational effects; it does not confirm that such effects followed. The source material also relays a security leader’s expectation of a “vulnerability tsunami” within months and an estimate of roughly 18 months to hold on. Those are attributed opinions, not established timelines or quantified forecasts. The reports provide no common measurement showing how much AI has increased the overall rate of successful attacks.

Investigations and Patch Work Continue

ConnectWise said it had suspended the affected services while investigating; further findings would be needed to clarify the scope and any customer impact. Organizations using the service can review vendor updates, check account and application access logs, and follow their incident-response procedures, without assuming that an attempted login means a breach.

For industrial operators, the agencies’ advisory sets out immediate checks: confirm PLC inventories and firmware status, prevent direct internet access, restrict remote connections, and monitor controller traffic. For companies more broadly, the documented activity strengthens the case for reviewing third-party application permissions, enforcing phishing-resistant authentication where available, and setting patch priorities before new vulnerability reports arrive. OpenAI’s forecast of more AI-found flaws makes patch capacity a relevant concern, but the scale and timing of any increase remain unknown.

Key Questions

What did US agencies report about AI and industrial systems?

A joint advisory said actors used AI-generated scripts against internet-accessible Siemens S7 PLCs. The agencies described reconnaissance and capability development and warned the activity could precede operational effects; the advisory does not confirm that industrial operations were disrupted.

Did the Microsoft report confirm that customer data was stolen?

No. Microsoft said most of the targeting failed and cautioned that an affected app or successful sign-in does not necessarily show that data was accessed. ConnectWise said existing backup archives showed no indication of compromise while it investigated.

What does the reported 2,250-tenant figure mean?

Microsoft said the actor used credentials associated with the cloud backup app to attempt sign-ins across more than 2,250 tenants. It is a count of targeted tenants, not a count of confirmed breaches.

What should organizations do now?

Review asset inventories and patch deadlines, protect remote access with strong authentication, check third-party application permissions, separate IT and operational networks, and practise incident response. Operators of Siemens S7 systems should also follow the joint agencies’ specific recommendations for PLC exposure and monitoring.

Is a coming “vulnerability tsunami” confirmed?

No. That phrase and the roughly 18-month estimate were presented as one security leader’s opinion. OpenAI separately forecast that AI-found critical vulnerabilities would increase sharply in coming months, but the source material does not provide a measured estimate of the scale or timing.

Source: ThorstenMeyerAI.com