An AI pentest is penetration testing that uses artificial intelligence to assist with offensive security tasks. The AI may help with repetitive work such as note-taking, log review, asset enumeration, pattern matching, or generating test ideas, while a human tester decides what is valid, safe, and worth escalating. It is best understood as a coverage amplifier, not a replacement for skilled judgment.
This matters because many weaknesses are subtle and context dependent. AI can help surface hidden vulnerabilities in web apps, APIs, or AI-driven systems, but it can also miss business logic flaws, chained abuse paths, or risks that require environmental context. In real attacks and defenses, similar techniques appear when adversaries automate reconnaissance or probe AI workflows with prompt injection and tool abuse, while defenders use AI-assisted testing to review more surface area, document findings, and check that connected tools obey scope and privilege limits.