Why Companies And Users Need To Rethink Cybersecurity In The AI Era

🔍 Read the full analysis: Why Companies And Users Need To Rethink Cybersecurity In The AI Era 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 Microsoft have reported separate cyber operations involving AI-generated scripts and an agentic AI framework, including activity aimed at industrial controllers and cloud tenants. The reports do not show that AI makes attacks routinely successful, but they document software carrying out familiar tasks at speed and have renewed questions about how quickly organizations can detect and contain intrusions.

A joint Cybersecurity Advisory published August 18 by the NSA, CISA, FBI, Department of Energy and Environmental Protection Agency said actors were using AI-generated scripts against internet-reachable Siemens S7 programmable logic controllers. The agencies described reconnaissance and capability development involving controllers used in sectors including manufacturing, energy, water and food and agriculture. They assessed the activity as likely preparation to cause operational effects, rather than reporting that such effects had occurred.

Microsoft Threat Intelligence reported October 21 that Storm-0940, which Microsoft describes as a China-state-aligned actor active since 2021, used the open-source ARTEX multi-agent penetration-testing framework in operations linked to a ConnectWise Cloud Backup compromise. Microsoft said the actor used credentials associated with the cloud backup application to attempt logins 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 failures per minute. The company said the majority of targeting was unsuccessful.

Microsoft also said that, in several attacks, the time from initial targeting to privileged cloud access was under three hours. It cautioned that an affected application or successful sign-in does not necessarily mean 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.

Speed Raises the Cost of Delays

The reports describe a shift in how quickly routine attack steps can be carried out, not evidence of an autonomous system that can reliably break into any target. Reconnaissance, password attempts and adapting scripts are established parts of cyber operations. Automation can let an operator attempt more of them, or keep them running for longer, while exploiting weak access controls or trusted connections.

That puts pressure on organizations to reduce the time between a vulnerability being disclosed and a fix being applied, and between suspicious activity and a response. The risks extend beyond a single company when attackers exploit a trusted supplier or application: customers may inherit exposure through access they previously granted. For industrial operators, internet-accessible controllers can also connect cyber incidents to physical processes, making network separation and careful remote-access controls important.

The same faster cycle may affect defenders. OpenAI said AI had helped find critical PHP vulnerabilities that were patched on October 21, and forecast that AI-discovered critical vulnerabilities would “increase sharply in the coming months.” That is a company forecast, not a measured prediction of how many flaws will appear or how many will be exploited. Even if discovery accelerates, organizations still need to verify, prioritize and install fixes.

Two Reports, Different Attack Paths

The August advisory concerns operational technology: programmable logic controllers that manage industrial equipment. The agencies said actors found exposed devices through internet scanning services and used scripts built with the open-source snap7 library, disguised as monitoring tools. They recommended inventorying controllers, updating firmware, keeping devices off the public internet, separating IT and operational networks, and requiring multifactor authentication for remote access.

Microsoft’s October account concerns a third-party cloud service and tenant access. The company said ARTEX can support a chain of tasks from reconnaissance to data exfiltration with limited human interaction. That description is Microsoft’s assessment of the framework and its use; it does not establish that every step in every incident was automated. Microsoft’s figures also distinguish attempts from outcomes: the failed logons were not successful access, and the company said most targeting did not succeed.

A separate 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 an agent searched GitHub after an initial exploit attempt failed. The examples are evidence of specific reported uses of AI tools, not proof that all attackers use them or that AI alone caused the compromises.

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

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

Impact and Forecasts Still Unsettled

The reports do not establish that the industrial activity caused operational disruption, damage or safety incidents. The agencies described reconnaissance and preparation, and their assessment that actors were likely preparing for operational effects remains an assessment. In the ConnectWise case, the extent of any customer data access is not established by the information provided here; Microsoft warned that successful sign-in does not necessarily mean data was accessed, while ConnectWise said its existing backup archives showed no indication of compromise.

It is also unclear how much of the reported work was performed by AI agents without human direction, how often comparable operations are occurring, or how many organizations face similar exposure. A security leader’s forecast, relayed in the source material, of a “vulnerability tsunami” within months and roughly 18 months to prepare is an expert opinion, not a measured trend. OpenAI’s prediction about vulnerability discovery is also forward-looking. Neither supplies a confirmed timetable for future attacks or disclosures.

Prioritize Controls and Follow Investigations

Organizations should review the immediate controls named in the advisory and reports: maintain an accurate inventory, set and track patch deadlines, keep industrial controllers off the public internet, segment operational technology from corporate systems, and require multifactor authentication for remote access. They should also review third-party applications and the permissions granted to them, monitor unusual sign-ins, and practice incident response so teams know how to revoke access and contain affected systems.

For individuals, the source material recommends starting with the email account, which often supports password resets and access to other services. Use a unique password and phishing-resistant multifactor authentication where available, review active sessions and connected applications, and act on security alerts through the service’s official site or app.

The next evidence will come from updates to the agencies’ advisory, Microsoft’s and ConnectWise’s investigation findings, and any further disclosure about customer or operational effects. OpenAI’s forecast can be checked against future vulnerability disclosures and patch activity, but no specific increase or timetable is confirmed here.

Key Questions

Does the reporting show that AI cyberattacks are routinely successful?

No. Microsoft said most of the targeting it described was unsuccessful, and it recorded more than 151,000 failed logons during one campaign. The reports document AI-assisted activity, not routine success across targets.

Was data stolen from ConnectWise customers?

The supplied reporting does not establish that customer data was accessed or stolen. 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 industrial operators do first?

Follow the joint advisory’s basic controls: inventory controllers, update firmware, prevent direct internet access, separate IT and operational networks, and require multifactor authentication for remote access. Operators should also monitor controller traffic and rehearse incident response.

Is a major wave of AI-discovered vulnerabilities confirmed?

No. OpenAI forecast that AI-found critical vulnerabilities would increase sharply in the coming months. The source material provides no future count or measured trend confirming that forecast.

Source: ThorstenMeyerAI.com