# AI-Enabled Penetration Testing: What CREST’s New Accreditation Means for Cybersecurity **Published on:** 2026-09-04T00:00:00.000Z **Author:** Denis Kucinic, VP of Operations Artificial intelligence is changing the way cybersecurity teams work. Security professionals are using AI to analyze information, automate repetitive tasks, generate code, accelerate research, and identify potential vulnerabilities. Penetration testing is no exception. But as AI becomes more deeply integrated into security testing, organizations face a new question: How can they know that their penetration testing provider is using AI responsibly? The Council of Registered Ethical Security Testers ([CREST](https://www.crest-approved.org/threat-led-penetration-testing-guidance-for-financial-services/)) is beginning to provide an answer. In 2026, CREST introduced an optional [AI-Enabled Penetration Testing annex](https://www.packetlabs.net/posts/crest-ai-charter/) to its Penetration Testing Accreditation Standard. The framework establishes independently assessable requirements around the governance, oversight, transparency, and responsible use of AI in penetration testing. Packetlabs is among the first organizations to achieve the new accreditation. For organizations considering AI-enabled penetration testing, the development offers an important lesson: the question isn't whether a provider uses AI. It's whether AI is being used in a way that improves testing without compromising human judgment, security, or client trust. ## What is AI-Enabled Penetration Testing? AI-enabled penetration testing refers to the use of artificial intelligence within a broader professional penetration testing methodology. AI assists penetration testers with activities such as: * Reconnaissance and information gathering * Data analysis * Vulnerability research * Code generation * Attack-path analysis * Identifying relationships between findings * Automating repetitive tasks * Organizing and summarizing technical information However, AI-enabled testing should not be confused with automated vulnerability scanning. A vulnerability scanner may identify a known weakness based on predefined signatures or behaviors. A penetration tester goes further by determining whether that weakness can actually be exploited, how it could be chained with other vulnerabilities, and what an attacker could ultimately accomplish. AI can support that process, but it does not eliminate the need for human expertise. That distinction is central to responsible [AI-enabled penetration testing](https://www.crest-approved.org/ai-charter/). ## Why Did CREST Introduce an AI Pentesting Accreditation? AI adoption is accelerating throughout cybersecurity. As more providers incorporate AI into their services, customers need a way to evaluate whether that technology is being used appropriately. CREST's AI-Enabled Penetration Testing accreditation addresses this challenge by establishing requirements specifically for providers using AI as part of penetration testing. The annex focuses on areas including: * AI governance * Oversight * Transparency * Responsible organizational use * Maintaining professional judgment * Preserving the quality and integrity of penetration testing CREST describes the framework as a way to help ensure that AI enhances professional judgment rather than undermining it. This is an important distinction. AI adoption should not become a race to automate as much of a penetration test as possible. Instead, cybersecurity providers should determine where AI adds value, and establish controls around its use. ## Packetlabs Among the First AI-Enabled Pentesting Providers At Packetlabs, we are thrilled to be part of CREST's first cohort of organizations to achieve the new AI-Enabled Penetration Testing accreditation. The accreditation builds on Packetlabs' existing CREST accreditation and SOC 2 Type II attestation. CREST's marketplace identifies Packetlabs as a penetration testing provider with a strong emphasis on manual testing and OSCP-minimum certified pentesters. Our approach at Packetlabs is deliberately human-led. The company reports that 95% of its penetration testing is manual-driven, with 100% OSCP-minimum certified staffing and zero outsourcing. The objective isn't to replace experienced ethical hackers with artificial intelligence. It's to give those testers better tools. ## How AI Augments Penetration Testers One of the biggest misconceptions surrounding AI in cybersecurity is that automation and expertise are competing approaches. They don't have to be. Security professionals have used automation for years. Vulnerability scanners, scripts, exploit frameworks, enumeration tools, and custom tooling all help testers work more efficiently. AI is another evolution of that toolkit. The difference is that modern AI can potentially assist with tasks that traditionally required more manual analysis. For example, AI may help a tester identify relationships across large quantities of information or rapidly generate code that can then be reviewed and adapted by the tester. But the important step comes afterward: a qualified penetration tester still needs to determine whether the result is accurate, relevant, exploitable, and meaningful. ## Why Human-Led Penetration Testing Still Matters A real attacker doesn't necessarily exploit vulnerabilities individually. Attackers chain weaknesses. A low-severity vulnerability might provide initial access. A configuration issue might allow privilege escalation. Excessive permissions might enable lateral movement. Poor network segmentation might provide access to a critical system. Individually, those weaknesses may appear relatively minor. Together, they can result in a serious compromise. This is why effective penetration testing needs context and judgment. Packetlabs' human-led approach is built around identifying these real-world attack paths rather than relying solely on automated scanning. Our methodology covers infrastructure, cloud, identity, applications, and AI environments. ## What Responsible AI Use Looks Like in Pentesting Responsible AI use begins with understanding that customer data is not simply test material. During a penetration test, security professionals may encounter credentials, source code, customer information, proprietary business data, internal architecture, and other sensitive information. That creates significant considerations when AI is introduced into the workflow. Organizations should want clear answers to questions such as: ### What information can be provided to AI systems? A provider should understand what information is appropriate to process using AI and what should remain outside AI systems. ### Where is that information processed? Customers should understand whether sensitive testing information is being sent to external AI platforms or handled within controlled environments. ### Is customer information used to train models? This is an especially important question when evaluating external AI services. ### Who is responsible for validating AI-generated results? AI output should not automatically become a penetration testing finding. Qualified professionals should review and validate results before they are incorporated into a client deliverable. ### How is AI use documented? Organizations should be able to understand how AI fits into the provider's methodology and what controls govern its use. These considerations help distinguish responsible AI-enabled pentesting from simply adding an AI tool to an existing workflow. ## How Packetlabs Approaches AI in Penetration Testing Packetlabs' AI approach is consistent with its broader human-led testing philosophy. The company states that AI is used under direct tester oversight and that client data does not touch public AI services or train models. Packetlabs also says these commitments are documented in its statements of work. That creates an important model for AI adoption in cybersecurity: **AI can increase efficiency without removing accountability.** Rather than treating AI as an autonomous penetration tester, Packetlabs positions the technology as another tool available to its security professionals. The tester remains responsible for interpreting results and determining their significance. This approach also aligns with the purpose of CREST's AI accreditation: ensuring AI enhances professional judgment while maintaining the quality and integrity expected of accredited penetration testing services. ## AI-Enabled Pentesting vs. Automated Vulnerability Scanning The distinction between AI-enabled penetration testing and vulnerability scanning deserves particular attention. Automated vulnerability scanners are useful for identifying potential weaknesses at scale. But scanning is not equivalent to penetration testing. A scan may identify an outdated service, exposed port, missing security header, or potentially vulnerable software version. A penetration tester asks what happens next. Can the vulnerability actually be exploited? Can it be chained with another weakness? Can privileges be escalated? Can an attacker move laterally? Can sensitive information be accessed? Can the weakness ultimately affect the organization's operations, customers, or revenue? Packetlabs' positioning centers on answering those questions. Its penetration testing services are designed to uncover real-world attack paths and validate what attackers could actually accomplish. AI can help make parts of that investigation more efficient. It should not turn the engagement into a scan-and-report exercise. ## How AI Creates a New Attack Surface There is another side to the AI security conversation. Organizations are not only using AI to improve cybersecurity. They are also deploying AI-enabled applications, large language models (LLMs), AI agents, and other systems that create new attack surfaces. These systems need to be tested. AI and LLM applications can introduce risks including: * Prompt injection * Jailbreaking * Sensitive data exposure * Insecure integrations * Excessive permissions * Model manipulation * Inadequate output controls * API vulnerabilities * Insecure handling of confidential information Packetlabs offers AI and LLM penetration testing specifically to assess these risks. Its testing approach examines areas including prompt injection, data leakage, model abuse, and API exposure. This creates an important distinction between two related services. AI-enabled penetration testing focuses on how a penetration testing provider uses AI in delivering its service. [AI and LLM penetration testing](https://www.packetlabs.net/services/ai-llm-penetration-testing/) focuses on testing the security of an organization's AI-enabled systems. Organizations may ultimately need both. ## What Should Organizations Look for in an AI Pentesting Provider? The growing availability of AI-enabled security services makes provider selection more complicated. A company should not choose a pentesting provider simply because it advertises AI capabilities. Instead, ask how AI is being used. ### Look for independent assurance Accreditations can provide evidence that a provider's processes have been independently assessed. CREST's AI-Enabled Penetration Testing accreditation is specifically designed to provide assurance around responsible AI use within accredited penetration testing services. ### Ask about human oversight Find out whether qualified testers validate AI-generated output. AI should support security expertise rather than bypass it. ### Understand data handling Ask whether client data is submitted to public AI services, whether data is retained, and whether it can be used for model training. ### Evaluate the testing methodology Don't let "AI-powered" become a substitute for understanding what the provider actually tests. Look for evidence of manual testing, attack-path analysis, exploitation, validation, and business-impact assessment. ### Consider certifications and assurance CREST accreditation and SOC 2 Type II attestation can provide additional assurance when evaluating providers. Packetlabs holds both and emphasizes its manual-driven approach. ## Why AI Governance Will Become a Bigger Part of Pentesting AI adoption is unlikely to slow down. As organizations deploy more AI systems and security providers incorporate AI into their workflows, governance will become increasingly important. This could eventually create two parallel questions for security leaders: "Are we secure against AI-enabled threats?" and "Are our security providers using AI responsibly?" Both matter. A penetration testing engagement involves a high level of trust. Customers give testers access to systems and information specifically so those testers can simulate attacks. AI adds another layer to that relationship. Organizations therefore need confidence not only in the technical abilities of their pentesting provider, but also in how the provider handles AI. Independent standards can help establish that confidence. ## The Future of AI-Enabled Penetration Testing AI will likely become an increasingly common component of penetration testing. It can help security professionals work with larger quantities of information, accelerate repetitive tasks, investigate potential attack paths, and improve efficiency. But the best use of AI isn't necessarily the most automated one. The goal should be to combine technology with expertise. Packetlabs' approach reflects that philosophy: human-led penetration testing supported by targeted automation and AI under direct tester oversight. The company continues to emphasize manual testing as a central part of its methodology. CREST's new AI-Enabled Penetration Testing accreditation provides another layer of assurance around this approach. For organizations evaluating pentesting providers, the takeaway is straightforward: **Don't ask whether your pentesting provider uses AI. Ask how they use it, what safeguards surround it, and whether experienced security professionals remain accountable for the results.** That is the difference between AI as a marketing label and AI as a responsibly governed security capability. ## Conclusion AI is changing both sides of the cybersecurity equation. Organizations are deploying AI-enabled applications that need to be tested, while penetration testing providers are incorporating AI into their own workflows. Packetlabs approaches both challenges through human-led security testing, combining experienced ethical hackers with technology to uncover real-world attack paths. As one of the first organizations to achieve CREST's AI-Enabled Penetration Testing accreditation, Packetlabs can help organizations navigate an increasingly AI-driven threat landscape while maintaining the human expertise required to understand what attackers can actually accomplish. **Ready to go beyond automated scanning? Talk to Packetlabs about your penetration testing requirements.** ### Frequently Asked Questions #### What is AI-enabled penetration testing? AI-enabled penetration testing uses artificial intelligence to support parts of a professional penetration testing engagement. AI may assist with research, analysis, reconnaissance, coding, or other tasks while qualified penetration testers retain responsibility for validation and decision-making. #### What is CREST's AI-Enabled Penetration Testing accreditation? CREST's AI-Enabled Penetration Testing accreditation is an optional annex to its Penetration Testing Accreditation Standard. It establishes independently assessable requirements concerning the governance, oversight, transparency, and responsible use of AI in penetration testing. #### Is Packetlabs CREST-accredited for AI-enabled penetration testing? Yes. Packetlabs is among the first organizations to achieve CREST's AI-Enabled Penetration Testing accreditation. Packetlabs also holds CREST penetration testing accreditation and SOC 2 Type II attestation. #### Does AI replace penetration testers? No. AI can assist with certain penetration testing activities, but experienced security professionals remain important for validating results, understanding business context, identifying attack paths, and determining real-world impact. #### How does Packetlabs use AI in penetration testing? Packetlabs states that it uses AI under direct tester oversight. The company also states that client data does not touch public AI services or train models, with its commitments documented in statements of work. #### What is the difference between AI-enabled pentesting and AI/LLM pentesting? AI-enabled penetration testing describes the use of AI by the penetration testing provider while delivering security testing. AI/LLM penetration testing assesses the security of an organization's own AI systems, such as LLM applications and AI-enabled software. #### Why is human-led penetration testing important? Human-led testing provides the expertise and contextual judgment needed to determine whether vulnerabilities can be chained into meaningful attack paths. It also allows testers to validate AI- or tool-generated results rather than treating automated output as proof of risk. #### Should organizations use AI-enabled penetration testing? AI-enabled testing can provide value when AI is responsibly governed and used to augment qualified security professionals. Organizations should evaluate a provider's methodology, data-handling practices, human oversight, and independent assurance before selecting a service.