How To Respond When AI Speeds Up The Cyberattack Cycle

🔍 Read the full analysis: How To Respond When AI Speeds Up The Cyberattack Cycle on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

US agencies and cybersecurity firms have reported cases of AI-assisted cyber activity against industrial control systems and cloud services, while OpenAI says AI-discovered critical vulnerabilities are expected to rise. The reports describe software speeding up familiar attack steps, not evidence of an all-powerful automated attacker; they also document failed attempts and leave some effects unconfirmed. The immediate implication for organizations is to tighten access, patching and response practices as the pace of activity increases.

A joint advisory published August 18 by the NSA, CISA, FBI, Department of Energy and EPA says actors have used AI-generated scripts in activity targeting internet-reachable Siemens S7 programmable logic controllers. PLCs are industrial computers used to operate equipment and processes. The agencies say attackers scan for exposed controllers and use scripts disguised as monitoring tools, built with the open-source snap7 library. They assess the activity as likely preparation to cause operational effects, rather than reporting that such effects have occurred.

In an October 21 report, Microsoft Threat Intelligence linked Storm-0940, which it describes as a China-state-aligned actor, to use of ARTEX, an open-source agentic penetration-testing framework, in activity involving a compromise of ConnectWise Cloud Backup. Microsoft says the actor used credentials associated with the cloud 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 in a password-spraying campaign, peaking at 2,900 per minute; Microsoft says none of those attempts succeeded. In several other attacks, it said, 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 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. Separately, Palo Alto Networks’ Unit 42 reported an agent performing hundreds of hours of manual targeting analysis in minutes. OpenAI said AI had identified critical PHP vulnerabilities patched October 21 before attackers did, and forecast that such findings would increase sharply in coming months.

Faster Attacks Raise the Pace

The reports matter because automation can compress routine parts of an intrusion: scanning, trying credentials, adapting scripts and moving through access already granted to a trusted service. That can give defenders less time to detect unusual behavior and contain it. Microsoft’s account of privileged access in under three hours illustrates the potential speed in some cases, while the large number of failed logons shows that automation does not guarantee success.

The same shift may increase pressure on security teams from the defensive side. If AI-assisted research finds more software flaws, organizations may face more patches and less time to evaluate and install them. OpenAI’s prediction is a forecast, not a measured future rate. The practical case for action does not depend on that forecast: exposed systems, weak remote access and unreviewed third-party permissions are established risks that can be addressed now.

For industrial operators, internet-accessible controllers can connect digital intrusion attempts to physical processes. For cloud customers, a compromised vendor application can make use of access customers have already authorized. These are distinct risks, but both make asset inventories, carefully limited permissions and rehearsed response plans more valuable.

Three Reports, Different Evidence

The source material points to separate reports rather than one coordinated campaign. The government advisory covers activity against Siemens S7 controllers; Microsoft describes Storm-0940 and the ConnectWise Cloud Backup compromise; Unit 42’s July 30 report describes a separate actor using a DeepSeek model with the Hermes Agent framework. These cases should not be collapsed into a single group or attack.

The common development is the use of AI or agentic software to accelerate ordinary offensive tasks. Microsoft says ARTEX can support activity from reconnaissance through data exfiltration with limited human interaction. That description is of the framework’s capability; it does not establish that every stage occurred in each reported incident. Likewise, an AI-generated script is evidence of a tool being used, not proof that an attack succeeded.

The response guidance in the industrial advisory is largely foundational: inventory controllers, update firmware, keep PLCs off the public internet, separate operational technology from business networks, require multifactor authentication for remote access and monitor control-system traffic. Microsoft’s report also highlights the need to examine third-party applications and the access they hold. These measures address the access paths described in the reporting without assuming every organization faces the same threat.

“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 Open

The available reports do not establish that the industrial-controller activity caused operational disruption. The agencies describe likely preparation for operational effects, but the source material does not confirm such effects. In the ConnectWise case, Microsoft says a successful sign-in does not necessarily mean data was accessed, while ConnectWise says it found no indication that existing backup archives were compromised. The investigation status and any later findings are not provided here.

It is also unclear how quickly AI-assisted attacks will spread or how much of the work in reported incidents was performed autonomously rather than directed by people. OpenAI’s expectation of sharply rising vulnerability findings is a forecast. A security leader’s prediction of a coming “vulnerability tsunami” and roughly 18 months to respond is an expert opinion, not a measured timeline or an established deadline.

Prioritize Controls and Follow-Ups

Organizations should first verify what they operate and expose: inventory internet-facing services, cloud applications and industrial controllers; remove unnecessary public access; and set clear deadlines for applying security updates. Operators of industrial systems should review the government advisory’s recommendations, including network separation, remote-access protections and monitoring for unexpected controller writes.

Cloud customers should review third-party app permissions, confirm whether service access is still needed and check sign-in logs for unusual patterns. Security teams can test whether alerts and escalation paths work under compressed timelines, then practise isolating affected systems and contacting vendors. Boards can ask for evidence of these basic controls and response exercises rather than treating AI adoption as a substitute for security work.

For individual users, securing the email account is a useful starting point: use a unique password, turn on multifactor authentication where available and review recovery methods and active sessions. Further updates on the ConnectWise investigation, government advisories and new vulnerability disclosures will clarify whether the reported activity led to additional confirmed impacts.

Key Questions

Does the reporting show that AI attacks are succeeding more often?

No. The reports show AI being used to speed attack activity, but Microsoft said most targeting in the activity it described was unsuccessful. The available material does not establish a general increase in attack success rates.

Were industrial systems disrupted?

The August advisory describes activity against internet-reachable Siemens S7 controllers and says agencies assess it as likely preparation for operational effects. The source material does not confirm that the activity disrupted industrial operations.

Was data stolen from ConnectWise customers?

The information provided does not confirm data theft. Microsoft said a successful sign-in does not necessarily indicate data access, and ConnectWise said existing backup archives showed no indication of compromise while it investigated.

What should organizations do first?

Start with an accurate inventory, prompt patching, restricted internet exposure, phishing-resistant sign-in and reviews of third-party application access. Organizations operating industrial systems should also separate operational technology from business networks and practise incident response.

Source: ThorstenMeyerAI.com