What To Do Now As AI Speeds Up Cyberattacks

🔍 Read the full analysis: What To Do Now As AI Speeds Up Cyberattacks on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get hardware and tech essentials 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 firms have reported attackers using AI-generated scripts and agentic software against industrial control systems and cloud services, offering documented examples of how automation can speed up familiar attack techniques. The reports do not show that every attempt succeeded: Microsoft said most of the targeting it observed was unsuccessful, while agencies warned that some activity against industrial systems could be preparation for operational effects.

A joint advisory published Aug. 18 by the NSA, CISA, FBI, Department of Energy and Environmental Protection Agency said actors were conducting reconnaissance and developing capabilities against internet-reachable Siemens S7 programmable logic controllers, or PLCs. These industrial computers control equipment such as pumps and production lines. The advisory says attackers used AI-generated exploitation scripts disguised as legitimate monitoring tools and built with the open-source snap7 library. The named devices span Siemens S7-200 through S7-1500 models, including safety controllers.

On Oct. 21, Microsoft Threat Intelligence reported that Storm-0940, which Microsoft describes as aligned with the Chinese state and active since 2021, used the open-source ARTEX agent framework in activity connected to the ConnectWise Cloud Backup compromise. Microsoft said the actor used credentials for the backup application to attempt logins across more than 2,250 tenants. From Oct. 10 to 14, it logged more than 151,000 failed sign-ins in a password-spraying campaign, with a peak of 2,900 failures per minute; Microsoft said none of those attempts succeeded.

Microsoft reported that in several attacks, the period from initial targeting to privileged cloud access was under three hours. It also cautioned that most targeting failed 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 while investigating.

Faster Attacks Raise the Defense Tempo

The reports point to a change in the speed and scale of routine attack steps, rather than evidence of an autonomous system capable of breaking any target. Scanning, password spraying, exploiting exposed services and abusing trusted applications are established methods. Automating parts of that work can let attackers run more attempts and shorten the time defenders have to identify suspicious activity.

That compressed response window matters particularly for industrial environments and cloud accounts. A compromised controller can affect physical processes, while privileged cloud access can expose systems or data if an attacker uses it successfully. The agencies’ advisory says the activity may be preparation for operational effects; that assessment is a warning about potential intent, not confirmation that physical disruption occurred.

Automation also increases pressure on defenders to patch quickly. OpenAI said AI helped find critical vulnerabilities in PHP that were patched on Oct. 21, and forecast that similar findings would rise sharply in coming months. That is a prediction from the company, not a measured estimate of future vulnerability volume. Faster discovery can help defenders, but it can also make timely inventory, prioritization and remediation more demanding.

Three Reports, Different Evidence

The industrial-control advisory is a joint government warning, not a claim that every named PLC has been compromised. It describes reconnaissance and script development against internet-exposed Siemens controllers, and says the activity could prepare attackers to cause operational effects. The agencies recommend keeping controllers off the public internet, updating firmware and separating operational technology from corporate IT.

Microsoft’s Oct. 21 account concerns a separate cloud supply-chain incident. It ties Storm-0940’s activity to a compromised ConnectWise Cloud Backup application and describes ARTEX as a multi-agent penetration-testing framework capable of handling stages from reconnaissance to data exfiltration with limited human involvement. The reported sign-in counts and access times describe Microsoft’s observations; they do not establish that all targeted tenants were breached.

A July 30 report from Palo Alto Networks’ Unit 42 described a Chinese-speaking actor using the DeepSeek model with the Hermes Agent framework, directed through Telegram. Unit 42 said that when an initial exploit failed, the agent searched GitHub for another approach. In a separate assessment, the cybersecurity company said an agent completed hundreds of hours of manual targeting analysis in minutes. Those observations illustrate the potential for faster work, but they do not establish how commonly such techniques are used across all attackers.

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

— Microsoft Threat Intelligence

What the Reports Do Not Establish

The available reports do not establish how many organizations suffered confirmed data theft, lasting access or operational disruption as a result of the described activity. Microsoft specifically said a successful sign-in or affected application does not necessarily mean data was accessed, and ConnectWise said it had found no indication of compromise in existing backup archives. The full scope and outcome of the investigation were not provided in the source material.

The government advisory describes reconnaissance and capability development and assesses that the activity is likely preparation for operational effects. It does not confirm that attackers caused disruption to US critical infrastructure. Nor do the reports quantify how often AI-assisted techniques are used compared with conventional tools, or how much AI changes overall attack success rates.

A security leader’s forecast that conditions could worsen before improving, including a possible “vulnerability tsunami” within months and an estimate of roughly 18 months to adapt, is an expert opinion relayed by the source article. It is not a measured forecast or agency assessment. The timing and scale of any increase remain uncertain.

Practical Steps and Open Investigations

For companies, the immediate response is to strengthen the controls named in the advisory and incident reporting rather than assume that purchasing AI tools is the answer. Maintain an accurate inventory of internet-facing assets and industrial controllers, set patch deadlines, and review third-party applications and the permissions they hold. Require phishing-resistant sign-in where available, apply multi-factor authentication to remote operational-technology access, and separate OT networks from corporate systems.

For PLCs, the joint agencies recommend removing internet exposure, blocking TCP port 102 at the perimeter, patching firmware, and monitoring S7 traffic for writes outside approved change windows. They also call for a demilitarized zone between IT and OT and, where suitable, unidirectional gateways for data-historian connections. Organizations should practise incident response so teams can recognize abnormal access and limit it quickly.

Individuals can begin with their email account: use a unique password, turn on multi-factor authentication, review recovery details and connected applications, and treat unexpected login or sharing prompts cautiously. Companies should watch for further findings from the ConnectWise investigation and updated guidance from government agencies. The reports provide no confirmed timeline for a broader wave of AI-assisted attacks, so preparedness should be based on the documented methods and each organization’s exposure.

Key Questions

Have AI-assisted attacks caused confirmed damage to industrial facilities?

The joint US advisory describes reconnaissance and AI-generated scripts targeting Siemens PLCs, and warns that the activity may be preparation for operational effects. The source material does not confirm physical damage or disruption.

Did attackers successfully access all 2,250 Microsoft-reported tenants?

No. Microsoft said the actor attempted sign-ins across more than 2,250 tenants, but also said most targeting was unsuccessful. It reported privileged cloud access in several attacks, not successful access to every tenant.

Does a successful sign-in mean data was stolen?

Not necessarily. Microsoft said an affected application or successful sign-in does not by itself indicate that data was accessed. ConnectWise said existing backup archives showed no indication of compromise while it investigated.

What should organizations prioritize now?

Start with an accurate asset inventory, timely patching, phishing-resistant sign-in, reviews of third-party app access, separation of IT and OT networks, and rehearsed incident response. For industrial systems, remove public internet exposure and monitor control-system traffic.

Is a major increase in AI-found vulnerabilities certain?

No. OpenAI forecast that such findings would rise sharply in coming months, but that is a company prediction, not a confirmed count or independently established timeline.

Source: ThorstenMeyerAI.com