What if cybercriminals could steal your company's information and then ask artificial intelligence which assets are worth attacking next?
That is the troubling capability behind VOLTA, a newly discovered Windows infostealer that uses AI to evaluate stolen information and prioritize victims by their potential financial value.
According to research published by Cisco Talos on October 31, 2026, VOLTA collects credentials, browser sessions, financial information, and other sensitive data before using commercial AI models to help criminals determine what might be most profitable.
The larger concern is not simply that criminals are using AI. It is that stolen information can now be analyzed with less human effort, potentially making targeted cybercrime more accessible.
Key Takeaway: VOLTA demonstrates how AI can make stolen information more actionable for cybercriminals. Organizations need to protect more than passwords. They must also restrict what unauthorized software can do on Windows endpoints before sensitive information is collected and transmitted.
VOLTA is a Windows information-stealing malware program with an unusual additional capability: It uses AI to analyze what it steals.
According to Cisco Talos, VOLTA can collect:
The malware can also use stolen authenticated sessions to search email receipts and transaction histories, helping identify a victim's address, financial activities, and digital assets.
Once the information is assembled, VOLTA can submit a structured report through OpenRouter to an AI model.
Its default selection is Anthropic's Claude Opus 4.6, with Moonshot AI's Kimi K2 and Google's Gemini 2.5 Flash configured as alternatives if a model refuses or fails to complete the assessment.
The AI is asked to evaluate the victim's assets, potential financial value, and security practices.
Importantly, the AI does not perform the initial theft. It analyzes the information after the malware has collected it.
Because criminals no longer need to manually investigate every compromised account to identify the most valuable opportunities.
Consider an employee whose Windows computer contains access to Microsoft 365, customer records, financial applications, and cloud administration tools.
A traditional infostealer might collect credentials and session information for later examination.
VOLTA adds another step. AI can organize that information and help the criminal identify which accounts, assets, or weaknesses deserve attention.
For a business, the consequences could include unauthorized access to customer information, business email compromise, financial fraud, and further attacks against connected systems.
The broader financial exposure is significant.
According to IBM's 2026 Cost of a Data Breach Report:
These figures do not measure VOLTA specifically. They illustrate the growing business exposure associated with AI-enabled cybercrime.
VOLTA's AI primarily reduces the work required to evaluate stolen information. It does not automatically make every stage of an attack faster or invisible.
Cisco Talos describes the innovation as replacing work that would normally require a human analyst.
There are other concerning capabilities.
VOLTA can delete its executable after running, potentially making post-incident investigation more difficult. It also communicates through established third-party services rather than relying exclusively on attacker-owned infrastructure.
For security teams, this presents a challenge.
A successful alert that arrives after credentials and session tokens have already been transmitted may not prevent the immediate damage.
Detection remains essential, but the timing matters.
The question becomes whether organizations can prevent unauthorized collection and access before the stolen information reaches the attacker.
Antivirus and Endpoint Detection and Response (EDR) can play an important role in identifying suspicious behavior, investigating incidents, and containing compromised systems.
However, their effectiveness depends on the capabilities, configuration, and enforcement features of the individual product.
VOLTA illustrates why organizations should not rely solely on recognizing malicious software.
Attackers can change malware code, file names, and delivery techniques. AI may make some of these variations easier to produce.
But changing the malware does not necessarily change what it must do to succeed.
It still needs to execute, access sensitive information, interact with applications, and transmit stolen data.
What if you did not have to detect the attack in order to stop it?
That is where prevention-focused endpoint controls become important.
Isolation and Containment can restrict the endpoint behaviors an attacker needs, even when the specific malware has not been identified.
Think of it as limiting what software is permitted to do rather than assuming every running application should have unrestricted access.
Appropriate endpoint controls can restrict unauthorized execution, constrain trusted application behavior, and limit access to sensitive system resources.
This is especially relevant to VOLTA.
Cisco Talos identified a technique involving unauthorized headless Chrome execution to access information protected by Chrome's App-Bound Encryption. Researchers specifically recommend application controls that restrict which processes may launch Chrome in that manner.
This is a practical example of prevention based on behavior rather than malware identification.
AppGuard is a proven endpoint protection solution with more than a decade of production history focused on prevention through Isolation and Containment.
Its approach is designed to restrict unauthorized Windows endpoint behavior without requiring the software to first identify an attack as malicious.
That does not establish that AppGuard blocks every VOLTA technique. Individual protections and configurations must be validated against the relevant attack behaviors.
The larger lesson is straightforward:
The attack may be new. The actions it needs to perform on a Windows endpoint often are not.
VOLTA provides an opportunity to review whether your organization's security strategy can prevent credential theft, not simply detect its aftermath.
These steps address risks well beyond VOLTA, including conventional infostealers, credential-stealing malware, and emerging AI-assisted attacks.
Both, but VOLTA highlights an important distinction.
The malware's primary activities remain familiar: collecting credentials, accessing authenticated sessions, gathering sensitive information, and transmitting stolen data.
What has changed is the ability to analyze the results automatically and determine which stolen assets may be worth exploiting.
For business leaders, this raises an important question:
Are we spending too much effort trying to recognize every new attack and too little effort restricting what an attacker can do after gaining access to a Windows endpoint?
Detect and Respond remains necessary.
But prevention should not depend exclusively on whether a security tool has previously encountered a particular threat.
Unknown does not automatically mean unstoppable.
Organizations that combine detection, strong identity security, and endpoint Isolation and Containment can reduce the opportunities attackers have to turn a compromised computer into a broader business incident.
To explore the prevention-focused approach, visit the CHIPS AppGuard resources and learn how restricting unauthorized endpoint behavior can complement existing security investments.