How To Strengthen Your Defenses As AI Speeds Up Cyberattacks

🔍 Read the full analysis: How To Strengthen Your Defenses As AI Speeds Up Cyberattacks on ThorstenMeyerAI.com

Before you orderOffer from Amazon

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

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 researchers have reported attackers using AI-generated scripts and software agents in activity against industrial controllers and cloud services, adding evidence that AI can speed up familiar attack techniques. The reports describe attempted intrusions and, in Microsoft’s account, some cases of privileged cloud access in under three hours; they do not establish that every targeted organization suffered data theft or operational damage.

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-reachable Siemens S7 programmable logic controllers, or PLCs. These industrial computers can manage equipment such as pumps and valves. The agencies say the activity included reconnaissance and capability development and assess that it may be preparation for operational effects. They named manufacturing, energy, water, chemical, food and agriculture, and commercial facilities among the sectors targeted.

Microsoft Threat Intelligence reported October 21 that Storm-0940, which Microsoft describes as aligned with the Chinese state, used the open-source ARTEX agentic framework in activity connected to a ConnectWise Cloud Backup compromise. Microsoft says the actor used credentials associated with the backup application to attempt sign-ins across more than 2,250 tenants. Between October 10 and 14, Microsoft recorded more than 151,000 failed logons, with a peak of 2,900 per minute; none of those password-spray attempts succeeded. In several attacks, Microsoft said, the interval from initial targeting to privileged cloud access was under three hours.

The accounts carry important limits. Microsoft said most targeting was unsuccessful and that 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 and said it suspended Cloud Backup, SaaS Security and Cloud Management during its investigation. Separately, Unit 42’s July 30 report described a Chinese-speaking actor using the DeepSeek model with the Hermes Agent framework, directed through Telegram. The supplied report excerpt does not include the full sequence or outcome, so those details cannot be established here.

Faster Attacks Put Pressure on Defenders

The reports suggest AI’s near-term security effect is less about novel, autonomous attacks than about speed and scale. Reconnaissance, script adaptation and repeated login attempts are familiar methods; software agents can perform parts of that work for longer and faster than a human operator. Microsoft says ARTEX can operate across stages from reconnaissance to data exfiltration with limited human interaction, but that description of a framework’s capabilities does not prove every capability was used successfully in these cases.

That pace can narrow the time available to detect suspicious activity and respond. A route through a trusted third-party application can also extend exposure beyond the original vendor, which is why organizations need to review application permissions and access rather than treating a vendor’s compromise as isolated. At the same time, the large number of failed logons in Microsoft’s report is a reminder that volume alone is not proof of a breach.

OpenAI has said AI identified critical PHP vulnerabilities patched on October 21 before attackers did, and predicted that such findings will rise sharply in the coming months. That is a company forecast, not a measured estimate of future attacks. If more flaws are found and fixed quickly, defenders may face greater pressure to assess patches and deploy them promptly; the supplied information does not quantify the likely scale or pace.

Three Reports, Different Evidence

The August 18 government advisory concerns industrial control systems, while Microsoft’s October 21 report focuses on cloud access and a compromised backup application’s trust relationships. Unit 42’s July 30 report describes an AI model connected to an agent framework. These are separate accounts, not evidence of one coordinated campaign or a single tool used across all incidents.

The agencies say actors found exposed Siemens PLCs through internet scanning services and used AI-generated scripts disguised as monitoring tools, built with the open-source snap7 library. They characterize the activity as an evolution that reduces the technical expertise and time required. Their recommendations include taking PLCs off the public internet, maintaining firmware, separating operational technology from corporate IT, requiring multifactor authentication for remote access and monitoring for unauthorized changes.

Microsoft’s account also distinguishes attempted activity from confirmed access. It attributes the Cloud Backup activity to Storm-0940, but its figures include failed attempts, and its report cautions that even a successful sign-in does not establish data access. A separate expert forecast cited by ThorstenMeyerAI.com predicts a coming “vulnerability tsunami” and estimates roughly 18 months of pressure. That is an individual’s opinion, not a measured forecast or a government assessment.

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

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

Impact and Forecasts Remain Unsettled

The government advisory describes activity and assesses that it may be preparation for operational effects, but the supplied material does not confirm that attackers caused damage to industrial processes. Microsoft reported rapid access in several attacks while also saying most targeting failed; it did not establish that all affected tenants experienced data access or theft. ConnectWise’s statement about its archives is a company finding, and the investigation’s full results are not provided here.

It is also unclear how often AI-generated techniques are being used across the wider threat landscape, how much they improve attackers’ success rates, or whether the examples represent a lasting shift. Predictions of a sharp rise in vulnerabilities or a near-term “tsunami” remain forecasts, not confirmed outcomes. The available reports do not provide a common measurement that would allow a direct comparison of attack rates before and after AI use.

Prioritize Controls and Incident Readiness

Organizations should start with controls that address the documented paths: inventory internet-facing systems and industrial controllers, patch firmware and software, and remove unnecessary public access. For operational technology, the joint advisory calls for separation from corporate networks, multifactor authentication on remote access, and monitoring of control-system traffic for writes outside approved maintenance windows. It also recommends perimeter controls for PLC traffic and one-way gateways for historian connections where appropriate.

For cloud environments, security teams should review third-party application permissions, credentials and tenant access, and investigate unusual sign-in patterns. Phishing-resistant authentication, prompt access revocation when a vendor incident emerges, and rehearsed response procedures can reduce reliance on improvised decisions during an incident. Leaders should ask for evidence of these practices rather than treating adoption of an AI security product as a substitute.

Users can begin by securing their primary email account: use a unique password, enable multifactor authentication, review recovery methods and remove unfamiliar connected applications. The next milestones are the outcome of ConnectWise’s investigation, further technical updates from Microsoft and the agencies, and whether future vulnerability reports support OpenAI’s forecast. Until then, confirmed incidents and predictions should remain distinct.

Key Questions

Does the reporting show that AI attacks succeeded?

Some activity reached privileged cloud access, according to Microsoft, but its report says most targeting was unsuccessful. The supplied reports do not confirm that the PLC activity caused operational damage, and a successful sign-in does not by itself prove data access.

What is ARTEX?

Microsoft describes ARTEX as an open-source, multi-agent penetration-testing framework that can support activity from reconnaissance through data exfiltration with limited human interaction. That description is about its capabilities; it does not establish that every capability was used in the reported activity.

Which organizations should act on the PLC advisory?

The advisory names organizations using Siemens S7 PLCs, especially in manufacturing, energy, water, chemical, food and agriculture, and commercial facilities. Operators should check whether controllers are exposed to the internet and apply the advisory’s network, authentication and monitoring recommendations.

What can individuals do now?

Secure the email account used to reset other passwords: set a unique password, turn on multifactor authentication, review account recovery options and remove connected apps that are no longer needed. These steps do not eliminate risk, but they reduce common ways an account can be taken over.

Source: ThorstenMeyerAI.com