How To Prepare For Cyberattacks That AI Can Accelerate

🔍 Read the full analysis: How To Prepare For Cyberattacks That AI Can Accelerate on ThorstenMeyerAI.com

Age 18–24?Offer from Amazon

Prime made for students and young adults

• Fast, free delivery for dorm and study essentials
• Prime Video and Amazon Music included
• Member-only deals

Try Prime for Young Adults →Free trial for eligible 18–24 year olds

As an affiliate, we earn on qualifying purchases.Hide

US agencies and cybersecurity researchers have reported attackers using AI-generated scripts against industrial controllers and an agent framework in attacks involving a cloud backup app, raising concerns that automation can speed up familiar intrusion methods. The reports document activity, but also show that many attempts failed; they do not establish that AI has made attacks consistently successful.

A joint advisory published August 18 by the NSA, CISA, FBI, Department of Energy and EPA says actors are conducting reconnaissance and developing capabilities against internet-accessible Siemens S7 programmable logic controllers, used to operate industrial equipment. The agencies say attackers found exposed controllers through internet scanning services and used AI-generated exploitation scripts disguised as monitoring tools. They assess the activity as likely preparation to cause operational effects, not confirmation that such effects have occurred.

In an October 21 report, Microsoft Threat Intelligence said the China-state-aligned actor Storm-0940 used ARTEX, an open-source agentic penetration-testing framework, in activity linked by Microsoft to the ConnectWise Cloud Backup compromise. Microsoft said ARTEX was used to target more than 2,250 tenants after the actor obtained credentials for the backup app. A password-spray campaign recorded more than 151,000 failed sign-ins from October 10 to 14, with a peak of 2,900 failures per minute. Microsoft said none of those attempts succeeded. In several attacks, it said, the path from initial targeting to privileged cloud access took less than three hours.

Microsoft cautioned that most of the targeting was unsuccessful and that a successful sign-in or affected app does not by itself show that data was accessed. ConnectWise said existing backup archives showed no indication of compromise and said it suspended Cloud Backup, SaaS Security and Cloud Management during its investigation. Separately, OpenAI said AI had identified critical PHP vulnerabilities before they were patched on October 21 and forecast that AI-found vulnerabilities would increase sharply in coming months. That is a forecast, not a measured estimate of how many flaws will emerge.

Faster Attacks Raise the Response Bar

The reports point to a change in attack economics, rather than proof of a new kind of infallible attacker. Automated tools can repeat reconnaissance, credential attempts and parts of an intrusion process at machine speed. Microsoft says agent systems can work around the clock; Palo Alto Networks’ Unit 42 reported seeing an agent complete hundreds of hours of manual targeting analysis in minutes. Those are attributed observations, not a guarantee that every automated operation will succeed.

For organizations, faster activity can compress the time between a weak point being found and defenders needing to respond. A cloud account with broad permissions or an internet-exposed industrial controller can turn a basic security gap into a higher-impact risk. At the same time, AI-assisted vulnerability discovery may lead to more flaws being identified and patched, adding pressure on teams to triage updates promptly. The practical implication is to strengthen existing controls, not to assume that buying an AI security product alone will address the problem.

For users, a compromised email account can provide a route to password resets, financial fraud or access to workplace systems. Securing email with a unique password and phishing-resistant multi-factor authentication, where available, is a useful personal step. Organizations should also review third-party apps that hold delegated access, since attackers may try to exploit trust already granted to a service.

Three Reports, Different Attack Paths

The August advisory concerns operational technology: Siemens S7-200 through S7-1500 controllers, including safety controllers, used in sectors such as manufacturing, energy, water, chemicals, food and agriculture, and commercial facilities. The agencies recommend inventorying controllers, applying firmware updates, removing direct internet exposure, separating information technology from operational technology, requiring multi-factor authentication for remote access and monitoring controller traffic for unexpected writes.

The Microsoft account describes a different route: an attack involving a trusted software provider and access to customer tenants. ConnectWise disclosed its Cloud Backup compromise on October 16; Microsoft published its account five days later. Microsoft attributes the activity to Storm-0940, which it says has been active since 2021. Its report does not say that all targeted tenants were breached or that data was taken from every account where a sign-in succeeded.

A July 30 report from Unit 42 describes a separate case involving a Chinese-speaking actor who connected the DeepSeek model to the Hermes Agent framework and directed it through Telegram. The available source material says the agent searched GitHub after its first exploit attempt failed, but does not provide the rest of that case’s outcome. These reports describe distinct actors and methods; they should not be treated as evidence of one coordinated campaign.

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

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

Success and Impact Remain Unclear

The public reports do not establish how many industrial controllers were compromised or whether attackers caused physical or operational disruption. The US agencies assess that activity is likely preparation for operational effects, but the supplied advisory account does not confirm that such effects took place.

Microsoft’s report also leaves key incident details unresolved. It says most targeting failed and warns that an affected app or successful sign-in does not necessarily mean data was accessed. ConnectWise reported no indication of compromise in existing backup archives, but the investigation status and the full scope of any customer impact are not specified in the source material. OpenAI’s expected rise in AI-found vulnerabilities likewise has no quantified forecast or stated measurement window here.

A security leader quoted in the source material predicted a near-term “vulnerability tsunami” and said organizations might have roughly 18 months to prepare. That is an individual’s expert opinion, not a measured timeline or an official warning. The extent to which AI changes overall attack success rates remains uncertain.

Prioritize Controls and Incident Readiness

Organizations can act on the documented risks without waiting for a precise forecast. Security teams should maintain an accurate inventory of internet-facing assets, set patch deadlines based on exposure and severity, and confirm that industrial controllers are not directly reachable from the public internet. For operational technology, the agency advisory recommends separating IT and OT networks, using a demilitarized zone where appropriate, restricting remote access with multi-factor authentication and monitoring for unauthorized controller writes.

Cloud and identity teams should review third-party applications and the permissions granted to them, remove access that is no longer needed, and alert on unusual sign-in patterns. Where possible, use phishing-resistant authentication for privileged accounts. Boards and executives should ask whether these basic controls are in place, how quickly critical patches are applied and whether staff have practised responding to a compromised supplier or cloud account.

The next concrete developments are the outcome of ConnectWise’s investigation, any additional advisories from US agencies, and evidence about whether AI-assisted vulnerability discovery changes patch volumes. Until those arrive, organizations should treat the reported incidents as documented examples of faster automation, while keeping forecasts distinct from confirmed outcomes.

Key Questions

Have AI-driven attacks been confirmed?

AI-assisted activity has been documented in the reports cited here. US agencies say actors used AI-generated scripts against Siemens controllers, while Microsoft says Storm-0940 used the ARTEX framework. These reports do not show that AI makes attacks reliably successful.

Did the Microsoft-reported activity expose customer data?

The supplied reporting does not confirm that customer data was accessed. Microsoft said a successful sign-in or an affected app does not necessarily mean data was reached. ConnectWise said existing backup archives showed no indication of compromise while it investigated.

What should companies do first?

Start by identifying exposed systems and accounts, applying critical patches, removing unnecessary third-party app permissions and protecting remote access with multi-factor authentication. Organizations with industrial systems should check that controllers are not reachable from the internet and that IT and OT networks are appropriately separated.

Is a coming vulnerability surge certain?

No. OpenAI said AI-found critical vulnerabilities are expected to increase sharply in coming months, but that is a forecast. The source material provides no quantified estimate or measurement window to confirm the scale of any increase.

Source: ThorstenMeyerAI.com