AI Has Made Cyberattacks Faster. Here’s How To Respond

🔍 Read the full analysis: AI Has Made Cyberattacks Faster. Here’s How To Respond on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

• 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 Microsoft have reported documented cases of AI-assisted cyber activity against industrial control systems and cloud services, describing how automation can speed up reconnaissance, exploit attempts and account targeting. The reports show attacks involving internet-reachable Siemens controllers and a campaign that Microsoft says targeted more than 2,250 customer tenants, although most of the campaign’s attempts were unsuccessful.

A joint advisory published August 18 by the NSA, CISA, FBI, Department of Energy and Environmental Protection Agency says attackers used AI-generated scripts against Siemens S7 programmable logic controllers, or PLCs, in US critical infrastructure. PLCs are industrial computers that control equipment such as pumps and production lines. The agencies said actors found exposed devices through internet-scanning services and used scripts built on the open-source snap7 library, disguised as monitoring tools. They assessed the activity as likely preparation for operational effects, not proof that such effects had occurred.

In an October 21 report, Microsoft Threat Intelligence linked China-state-aligned actor Storm-0940 to use of ARTEX, an open-source agentic penetration-testing framework, in activity related to the ConnectWise Cloud Backup compromise. Microsoft said the actor attempted sign-ins across more than 2,250 tenants using credentials for the cloud backup application. Between October 10 and 14, Microsoft recorded more than 151,000 failed logons, peaking at 2,900 per minute; it said none of those attempts succeeded. In several attacks, Microsoft said, the period 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 customer data was accessed. ConnectWise, which disclosed the compromise on October 16, said existing backup archives showed no indication of compromise and said it suspended Cloud Backup, SaaS Security and Cloud Management while investigating. Separately, OpenAI said AI had found critical PHP vulnerabilities that were patched October 21, and forecast that similar findings would increase sharply in the coming months. That is a forecast, not a measured estimate of future incidents.

Automation Shortens the Attack Window

The reports matter because they describe ordinary attack techniques being carried out at machine speed, rather than evidence of an all-powerful or autonomous attacker. Automated systems can keep scanning, testing credentials or adapting scripts without the time limits faced by a human operator. Microsoft said agents can work around the clock; Palo Alto Networks’ Unit 42 described one system completing hundreds of hours of manual targeting analysis in minutes. These are company descriptions of particular activity, not a universal measure of how much faster every attack has become.

For defenders, a faster campaign can compress the time available to detect suspicious access and respond before attackers reach privileged accounts. The industrial advisory also highlights a physical risk: exposed controllers may govern processes in energy, water, manufacturing and other facilities. The agencies recommend reducing exposure and monitoring changes to controller traffic. Their warning that activity may be preparation for operational effects should not be read as confirmation that infrastructure was disrupted.

AI-assisted vulnerability discovery also cuts both ways. OpenAI’s account of vulnerabilities found before they were patched suggests AI can help identify flaws for remediation, while its forecast of sharply rising discoveries points to increased pressure on organizations to assess and apply fixes. The immediate practical implication is not that every company needs to buy an AI security product. It is that asset inventories, timely patching, strong authentication and practiced response plans become harder to postpone when attackers can move quickly.

Three Reports, Different Evidence

The government advisory, Microsoft report and OpenAI statement address different parts of the threat picture. The August advisory concerns AI-generated scripts and exposed industrial controllers. Microsoft’s October report describes an agent framework used in a campaign involving a compromised third-party application and cloud tenants. OpenAI’s statement concerns AI finding software vulnerabilities before attackers did. Together, the sources support the conclusion that AI is appearing in both offensive activity and defensive research; they do not establish that the same actors, tools or methods connect all three developments.

Microsoft said Storm-0940 has been active since 2021 and described ARTEX as capable of running tasks from reconnaissance to data exfiltration with limited human interaction. Those capabilities do not prove that every step in the reported campaign was autonomous. ConnectWise disclosed its compromise on October 16, and Microsoft published its account five days later. The difference between attempted access, successful sign-in and confirmed data theft remains important when interpreting the incident.

Advice in the industrial advisory is largely based on established network security practices: keep PLCs off the public internet, patch firmware, separate operational technology from corporate IT, require multifactor authentication for remote access and watch for unexpected controller writes. The source article also relays an unnamed security leader’s prediction of a coming “vulnerability tsunami” and an approximately 18-month period to respond. That is an individual’s forecast, not an agency finding or a measured timeline, and it should not be treated as confirmed evidence.

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

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

Access Does Not Prove Data Theft

The full impact of the ConnectWise-related activity remains unclear from the material reported here. Microsoft said successful sign-ins do not necessarily mean data was accessed, while ConnectWise said its existing backup archives showed no indication of compromise. The available accounts do not establish whether any tenant data was taken or whether all affected customers were identified.

The government advisory describes activity against internet-reachable PLCs and assesses that it may be preparation for operational effects. It does not confirm that attackers caused physical disruption. Nor do the reports establish that AI was responsible for every stage of each operation, or quantify how much AI increased the success rate compared with non-AI methods. OpenAI’s expected rise in AI-discovered vulnerabilities and the unnamed leader’s timeline are forecasts, not verified future outcomes.

Investigations and Basic Defenses

ConnectWise said it had suspended the named services while investigating; the cited material does not provide a final incident finding or a date for resuming them. Customers using the affected services should follow direct updates from ConnectWise and assess their own access logs and backup environments rather than assume that either compromise or safety has been established for every tenant.

For organizations with industrial systems, the agencies’ immediate recommendations include inventorying controllers, applying firmware updates, blocking internet access to PLCs, separating IT and operational networks and requiring multifactor authentication for remote access. They also recommend blocking TCP port 102 at network perimeters where appropriate and monitoring S7 traffic for writes outside approved change windows. Companies using third-party cloud applications can review granted permissions and credentials, restrict access where possible, and prepare to revoke an app’s access if an investigation calls for it.

For individual users, securing the email account is a useful starting point: use a unique password, turn on multifactor authentication and review recovery methods and active sessions. More broadly, the next developments to watch are findings from the ConnectWise investigation, any further government guidance on exposed controllers, and evidence showing how often AI-assisted activity leads to confirmed access or operational impact.

Key Questions

What AI-assisted cyber activity has been reported?

US agencies said attackers used AI-generated scripts against Siemens S7 industrial controllers. Microsoft reported that Storm-0940 used the ARTEX framework in activity targeting cloud tenants through a compromised backup application.

Does the Microsoft report confirm that customer data was stolen?

No. Microsoft said a successful sign-in or an affected application does not necessarily indicate data access. ConnectWise said its existing backup archives showed no indication of compromise while it investigated.

Were the reported attacks successful?

Not all were. Microsoft said most of the targeting was unsuccessful and that none of the more than 151,000 failed logons it recorded during the specified October 10–14 campaign succeeded. It also described privileged cloud access in under three hours in several attacks.

What should organizations do about exposed industrial controllers?

The joint agency advisory recommends inventorying PLCs, patching firmware, keeping controllers off the public internet, separating IT and operational technology networks, requiring multifactor authentication for remote OT access and monitoring controller traffic for unauthorized writes.

Is a surge in AI-discovered vulnerabilities confirmed?

No. OpenAI said it expects AI-found critical vulnerabilities to increase sharply in the coming months. That is a forecast; the source material does not provide a measured future total or a confirmed timeline.

Source: ThorstenMeyerAI.com