AI Is Changing The Pace Of Cyberattacks. Here’s What To Do

🔍 Read the full analysis: AI Is Changing The Pace Of Cyberattacks. Here’s What To Do on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

• 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

U.S. agencies and Microsoft have reported separate cases in which attackers used AI-generated scripts or an agentic framework in activity targeting industrial control systems and cloud accounts. The reports show how automation can speed up familiar attack methods, but they do not establish that most attempts succeeded or that AI alone caused the compromises.

A joint advisory published Aug. 18 by the NSA, CISA, FBI, Department of Energy and Environmental Protection Agency says actors used AI-generated exploitation scripts against internet-accessible Siemens S7 programmable logic controllers, or PLCs. These industrial computers help operate equipment such as pumps and production lines. The advisory says actors used scripts disguised as monitoring tools and built with the open-source snap7 library. It describes the activity as an active threat and assesses that it may be preparation for operational effects against critical infrastructure.

In an Oct. 21 report, Microsoft Threat Intelligence said the China-state-aligned actor it calls Storm-0940 used ARTEX, an open-source agentic penetration-testing framework, in activity linked to the ConnectWise Cloud Backup compromise. Microsoft said the actor tried to sign in to more than 2,250 tenants using credentials obtained from the cloud backup app. Between Oct. 10 and 14, Microsoft recorded more than 151,000 failed logons, with a peak of 2,900 failures per minute. It said none of those attempts succeeded. In several attacks, the time from initial targeting to privileged cloud access was under three hours.

Microsoft cautioned that most of the targeting was unsuccessful and that an affected app or successful sign-in does not by itself mean data was accessed. ConnectWise said existing backup archives showed no indication of compromise. It suspended Cloud Backup, SaaS Security and Cloud Management while investigating. Separately, OpenAI said AI had identified critical PHP vulnerabilities patched on Oct. 21 and forecast that AI-found critical vulnerabilities would “increase sharply in the coming months.” That is a forecast, not a measured rate of future discoveries.

Automation Compresses Attack Timelines

The reports matter because they show automation being applied to routine parts of cyberattacks: finding exposed systems, testing credentials and adapting tools. Shorter timelines can leave defenders less time to spot activity and restrict access, especially when a compromised vendor application already has permissions across customer environments.

That does not mean attackers have become infallible or that every AI-assisted attempt works. Microsoft said most targeting in the cases it described did not succeed, and its account of 151,000 failed logons makes clear that high volume is not the same as a successful breach. The practical concern is that automation can let an attacker make more attempts, or move between steps faster, while defenders still need to identify the activity and respond.

AI can also help find software flaws, according to OpenAI’s account of the PHP vulnerabilities. That may bring more fixes, but it can also increase the pressure on organizations to identify affected software and apply patches promptly. The immediate response should be grounded in established security controls, not an assumption that buying an AI product will solve the problem.

Three Reports, Different Evidence

The incidents described are not a single coordinated campaign. The August advisory concerns activity against Siemens industrial controllers; Microsoft’s October report concerns Storm-0940, ARTEX and a compromised cloud backup application. A separate Unit 42 report, published July 30, described a Chinese-speaking actor using the DeepSeek model through the Hermes Agent framework and steering it over Telegram. These accounts provide examples of AI-enabled or agent-assisted activity, but they do not establish how common such methods are across all cyberattacks.

The industrial advisory says exposed PLCs were found through internet-scanning services and warns that the targeted sectors include manufacturing, energy, water, chemicals, food and agriculture, and commercial facilities. It recommends keeping controllers off the public internet, inventorying devices, updating firmware, separating information technology and operational technology networks, and requiring multifactor authentication for remote access. For appropriate connections, it also recommends unidirectional gateways and monitoring S7 traffic for writes outside approved change windows.

Microsoft describes ARTEX as a multi-agent framework capable of running through steps from reconnaissance to data exfiltration with limited human involvement. In the reported activity, the app’s existing trust relationship mattered: credentials associated with the backup service were used to attempt access to customer tenants. Microsoft’s account and ConnectWise’s investigation statements should be read separately; neither confirms that customer data was accessed.

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

— Microsoft Threat Intelligence

Success and Scale Remain Unclear

The public reports do not establish how often AI is used in cyberattacks overall, or how much of the activity would have been possible with conventional automation. The agencies say the PLC activity may be preparation for operational effects; the advisory does not confirm that the activity caused physical disruption. Details about the full scope and consequences of the ConnectWise incident remain under investigation.

Microsoft said an affected application or a successful sign-in does not necessarily indicate that data was accessed. ConnectWise reported no indication of compromise in existing backup archives, but the available statements do not resolve whether all potentially affected customer environments have been reviewed. OpenAI’s prediction about a sharp increase in AI-found vulnerabilities is also a forecast rather than an observed future trend.

Immediate Steps for Operators

Organizations should begin by identifying internet-facing systems, cloud applications and operational technology assets, then confirm who can access them and what permissions third-party apps hold. For PLC operators, the agencies’ recommendations include removing public internet exposure, updating firmware, separating IT and OT networks, enabling multifactor authentication for remote access and monitoring for unauthorized control commands.

For cloud services, security teams can review third-party app access, revoke credentials that are no longer needed, use phishing-resistant sign-in methods where available, and set clear patch deadlines. Boards and executives can ask whether those controls are in place and whether incident-response plans have been practised. Individuals can start with their email accounts: secure them with multifactor authentication, use a unique password and review recovery settings.

The next concrete developments are the outcome of ConnectWise’s investigation and any further advisories from government agencies or threat researchers. Until those findings are available, the published reports support a measured conclusion: AI is appearing in documented attack activity, but basic access, patching and network controls remain the recommended response.

Key Questions

Did the reported AI-assisted attacks succeed?

Some activity reached privileged cloud access in several cases, according to Microsoft. But Microsoft said most targeting was unsuccessful and that a sign-in does not necessarily mean data was accessed. The government advisory describes PLC targeting but does not confirm operational disruption.

What should organizations do first?

Identify exposed systems and review access. The agencies recommend keeping PLCs off the public internet, patching firmware, separating IT and OT networks, requiring multifactor authentication for remote access and monitoring control-system traffic.

Does the reporting prove AI is causing more cyberattacks?

No. The reports document specific AI-assisted activity, but do not establish how common it is or prove an overall increase in attacks. OpenAI’s statement that AI-found critical vulnerabilities will rise sharply is a forecast.

Was customer data taken in the ConnectWise incident?

The available statements do not confirm that customer data was accessed. Microsoft said an affected app or successful sign-in does not necessarily indicate data access. ConnectWise said existing backup archives showed no indication of compromise while it investigated.

Source: ThorstenMeyerAI.com