/ BLOG

Will AI replace the ethical hacker during PCI DSS penetration testing?


AI finds vulnerabilities, analyzes code, checks for exposed services, and generates reports in minutes. For a company preparing for a PCI DSS audit, this is an opportunity to reduce costs and speed up the audit process.

However, when it comes to protecting payment data, speed isn’t the most important factor. Is an automated tool capable of independently verifying whether an attacker could infiltrate the Cardholder Data Environment (CDE)? Let’s take a closer look.

What requirements does PCI DSS set for penetration testing?


PCI DSS does not prohibit the use of AI or automated scanners. At the same time, Requirement 11.4 specifically governs penetration testing (pentests) and is not limited to running a vulnerability scanner. It requires a comprehensive internal and external penetration test conducted according to a clearly documented methodology.

During the test, the specialist not only identifies vulnerabilities but also verifies whether they can be exploited and assesses the potential impact on the company’s business processes.

Key requirements of the standard:

  • Internal and external penetration tests are conducted annually and following significant changes to the infrastructure or applications.
  • The testing covers the internal and external perimeter of the CDE within a scope agreed upon with the auditor, at the network and application levels.
  • Testing must be performed by a qualified internal specialist or an external contractor who is organizationally independent from the department that maintains these systems.
  • CDE segmentation must be tested separately to ensure that systems outside the scope of the test do not compromise its security.
  • Identified vulnerabilities are addressed in accordance with the risk assessment, and retesting is then conducted.

It is important to distinguish between requirements 11.3 and 11.4. Under requirement 11.3, scanning identifies known vulnerabilities, outdated software, and configuration errors. In contrast, a penetration test under requirement 11.4 verifies whether the identified weaknesses can be exploited in a real-world attack: to bypass security mechanisms, escalate privileges, or gain access to the CDE.

Automated tools and AI accelerate the search for potential vulnerabilities, analyze large volumes of data, and identify priority areas for testing. Their results serve as a useful data source for penetration testing but do not guarantee compliance with Requirement 11.4.

During testing, the specialist not only identifies vulnerabilities but also verifies whether they can be exploited and assesses the potential impact on the company’s business processes.

Why is the scope of the audit important?


The CDE is more than just a database of card records. The PCI DSS scope covers systems, processes, and technologies that store, process, or transmit cardholder data or sensitive authentication data. It includes components that are connected to the CDE or could impact its security:

  • authentication systems;
  • DNS and network services;
  • remote access tools;
  • CI/CD components and cloud infrastructure;
  • network security tools;
  • payment pages and related systems.

Automated tools alone are not sufficient to properly define the PCI DSS scope. It is necessary to consider the architecture, data flows, integrations, access models, tokenization, and the logic of the payment process.

If segmentation does not adequately isolate the CDE, the PCI DSS perimeter extends to additional systems and network segments.

Why Can’t AI Perform the Entire Penetration Test?


The idea of entrusting the entire penetration test to automated tools seemed logical. However, the experience of companies that have implemented this approach suggests otherwise.

According to the Cobalt AI and Pentesting Pulse Report 2026, the percentage of organizations that rely entirely on AI during testing has dropped from 29% to 9%. Instead, 47% of respondents opt for a hybrid model.

78% of those surveyed have encountered situations where automated tools missed critical vulnerabilities that experts later found manually. Therefore, in practice, AI is tasked with routine data collection, while the assessment of context and business risks is left to experts.

What Falls Outside the Scope of Automated Testing?


Despite the speed and efficiency of automated tools, some aspects of testing require the involvement of specialists.

1. Business logic of the payment process
Scanners are good at identifying individual vulnerabilities, but they do not allow for the evaluation of complex attack scenarios in the context of a specific business.

For example, a scanner will check the system and find no known CVEs—the system will be marked as secure in the report. However, an experienced specialist analyzes the system more deeply. They will change the order ID in an API request and see that access to someone else’s transaction has been granted. They’ll examine the authorization mechanism and find a flaw that allows them to elevate privileges to the level of an administrative account. This doesn’t mean the scanner missed a vulnerability: such scenarios go beyond its capabilities and require an analysis of the system’s operational logic and non-obvious attack vectors.

It is precisely the testing of such business scenarios that distinguishes a penetration test from automated scanning.

2. False Positives
Automated tools often generate false positives—that is, they flag vulnerabilities even when security measures are configured correctly. This increases the volume of audits and makes it harder to distinguish real risks from false positives.

A much more serious problem is complex attack scenarios where individual steps appear safe but, when combined, lead to a payment data breach. In the context of PCI DSS, this is particularly dangerous because a company sees a “clean” report from an automated scanner and mistakenly believes it is secure, while actual vulnerabilities remain undetected.

3. Attack Chains
The greatest risks arise when several minor system flaws combine to form a single attack scenario. For example, a configuration error grants access to an internal service; excessive permissions allow an attacker to compromise a service account; and the privileges of that account open the door to the CDE. Each of these issues requires attention, but it is their combination that transforms individual vulnerabilities into a full-fledged attack vector.

Automated tools detect only individual vulnerabilities, but linking them into a chain requires experience and an understanding of how system components are interconnected. A penetration tester manually models such scenarios based on the system’s architecture and how data flows between components.

4. Reporting to the QSA
To confirm compliance, a Qualified Security Auditor (QSA) reviews the scope of work, penetration test results, and adherence to the methodology. The documentation must include evidence of the systems tested, the methods used, the vulnerabilities identified, their remediation, and retesting.

An AI tool merely generates a list of identified vulnerabilities but does not explain the reasons for the chosen scope, the impact of the identified issues on the payment processing environment, or the methods for resolving them. This is the specialist’s responsibility. 

A Hybrid Approach in Practice


The main goal is to divide tasks between AI and an expert.

A penetration tester uses automation for the following tasks:

  • reconnaissance and inventory of the external attack surface;
  • initial scanning of large-scale networks;
  • static code analysis and detection of common vulnerabilities;
  • report generation;
  • preliminary prioritization of vulnerabilities prior to verification.

Automated tools perform standard checks, while the penetration tester investigates complex attack scenarios and analyzes business—

Preparing for a PCI DSS Penetration Test


Before selecting a contractor, clarify which systems the test will cover, which methods will be used, and to what extent they comply with the standard’s requirements:

CDE Scope

All systems, entry points, APIs, roles, and components that affect card data security are included.

Perimeter and Audit Levels

The internal and external perimeters, network and application levels, the CDE perimeter, and critical systems are covered.

Segmentation Testing

The use of segmentation to reduce the PCI DSS scope requires mandatory testing of its effectiveness.

Independence and Qualifications

The auditor has experience working with payment environments and is organizationally independent from the department that maintains the systems within the testing scope.

Expert Verification

The methodology includes verifying whether identified vulnerabilities can actually be exploited, analyzing complex scenarios, and manually verifying the results of automated tools.

Retesting

The scope of work includes mandatory retesting to confirm that identified vulnerabilities have been remediated.

Reporting for the QSA

The final report includes a justification of the scope, technical evidence, a risk assessment, and recommendations for remediation.

Human or Artificial Intelligence?


AI has become an important tool for penetration testers. It speeds up routine tasks, expands coverage, and helps identify issues that might otherwise go unnoticed due to the large volume of data. However, it cannot replace a specialist, especially during PCI DSS testing.

The standard requires that penetration tests be conducted by qualified specialists. An auditor needs a well-reasoned report, and a business needs an understanding of real risks—not just a list of vulnerabilities.

Don’t wait until weaknesses in your payment environment lead to an incident. Contact the ITS Red Team to identify vulnerabilities in a timely manner, assess potential risks, and determine priority security measures.

Need expert advice?