--- title: AI pentesting: what it is, and what it can actually find description: AI pentesting is explained as two distinct disciplines: using autonomous agents to run a penetration test against an organization's own apps and APIs, versus... canonical: https://daily.dev/posts/ai-pentesting-what-it-is-and-what-it-can-actually-find-w5c7pfldb twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI pentesting: what it is, and what it can actually find | daily.dev og:description: AI pentesting is explained as two distinct disciplines: using autonomous agents to run a penetration test against an organization's own apps and APIs, versus... og:url: https://daily.dev/posts/ai-pentesting-what-it-is-and-what-it-can-actually-find-w5c7pfldb og:image: https://api.daily.dev/og/posts/W5C7PFlDb.png og:image:alt: AI pentesting: what it is, and what it can actually find og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # AI pentesting: what it is, and what it can actually find **[Security Boulevard](https://daily.dev/sources/securityboulevard)** · 15 min read · 1 upvotes · 0 comments ## Summary AI pentesting is explained as two distinct disciplines: using autonomous agents to run a penetration test against an organization's own apps and APIs, versus testing an AI system's model, prompts and tooling for flaws like prompt injection. The piece argues that what an AI pentesting agent finds depends far more on its harness (a current attack-surface map, working credentials, multiple identities, and an enforced scope) than on which underlying model powers it, citing DARPA's AI Cyber Challenge, Google's Big Sleep/Project Zero SQLite find, and Google Threat Intelligence Group research on AI reasoning about business logic flaws. It also covers current limitations (non-determinism, reliability, accountability), what compliance assessors actually require, and heavily promotes Escape's Cascade and DAST products as built around these four inputs. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is the difference between AI pentesting and pentesting an AI system? They are two different jobs that share a confusing phrase. AI pentesting means an autonomous agent runs a penetration test against your own web apps and APIs, reproducing attacker behavior like reconnaissance and request chaining. Pentesting an AI system instead means assessing the model itself, its prompts, training data, retrieval layer, and the tools it can call, which is also called AI red teaming. _Teams sorting out which of these two testing jobs they actually need can track the distinction through coverage like this on daily.dev._ ### Does adding a second test identity actually find more vulnerabilities in an AI pentest? Yes, adding a second account to an AI pentesting agent typically surfaces 30 to 50 percent more issues, because an agent testing with only one identity has nothing to compare against and cannot detect broken object-level authorization. A benchmark that held the model constant but changed only the harness found four times as many real findings at a zero percent false-positive rate. _Developers comparing pentesting tool setups can follow benchmarks like this one on daily.dev before picking a vendor._ ### Can AI agents find vulnerabilities that traditional vulnerability scanners miss? Yes, AI agents can chain multi-step, identity-dependent flaws that scanners structurally cannot test, such as whether one user can reach another user's data, whether a business process can be completed out of order, or whether a bug chains into remote code execution. Google's Big Sleep agent found an exploitable stack buffer underflow in SQLite, described by Project Zero as the first public case of an AI agent finding a previously unknown exploitable memory-safety issue in widely used real-world software. _Engineers deciding whether agentic tools add real coverage beyond scanners can follow these findings on daily.dev._ ## Similar posts on daily.dev - [What Is AI Pentesting and How Does It Works?](https://daily.dev/posts/what-is-ai-pentesting-and-how-does-it-works--6qsoqaxod) · Snyk · 0 upvotes · 0 comments - [AI can find vulnerabilities. Humans find ways in](https://daily.dev/posts/ai-can-find-vulnerabilities-humans-find-ways-in-mi967qov8) · TechCentral · 0 upvotes · 0 comments - [AI pentesting tools are generating more findings than security teams can validate, new survey finds](https://daily.dev/posts/ai-pentesting-tools-are-generating-more-findings-than-security-teams-can-validate-new-survey-finds-am1p6nuyg) · IT Security Guru · 0 upvotes · 0 comments - [Jason Haddix: 90% of Pentests Will Be Done by AI](https://daily.dev/posts/jason-haddix-90-of-pentests-will-be-done-by-ai-fumwybxok) · Aikido Security · 0 upvotes · 0 comments - [AI Decline? Confidence Falls in Autonomous Penetration Testing](https://daily.dev/posts/ai-decline-confidence-falls-in-autonomous-penetration-testing-1pxtohmss) · Dark Reading · 1 upvotes · 0 comments --- Tags: [#security](https://daily.dev/tags/security), [#ai-agents](https://daily.dev/tags/ai-agents), [#vulnerability](https://daily.dev/tags/vulnerability), [#appsec](https://daily.dev/tags/appsec) [View this post on daily.dev](https://daily.dev/posts/ai-pentesting-what-it-is-and-what-it-can-actually-find-w5c7pfldb) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"AI pentesting: what it is, and what it can actually find","url":"https://daily.dev/posts/ai-pentesting-what-it-is-and-what-it-can-actually-find-w5c7pfldb","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-pentesting-what-it-is-and-what-it-can-actually-find-w5c7pfldb"},"datePublished":"2026-10-01T14:43:46.383Z","dateModified":"2026-10-01T14:44:17.463Z","description":"AI pentesting is explained as two distinct disciplines: using autonomous agents to run a penetration test against an organization's own apps and APIs, versus...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/d92ccab89a2927d30ca3904b0a6e5767?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/d92ccab89a2927d30ca3904b0a6e5767?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Security Boulevard","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Security Boulevard","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/3613c832180040de8d85bb29f74395be","url":"https://daily.dev/sources/securityboulevard"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-pentesting-what-it-is-and-what-it-can-actually-find-w5c7pfldb","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"security,ai-agents,vulnerability,appsec","timeRequired":"PT15M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Security Boulevard","item":"https://daily.dev/sources/securityboulevard"},{"@type":"ListItem","position":3,"name":"AI pentesting: what it is, and what it can actually find"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-pentesting-what-it-is-and-what-it-can-actually-find-w5c7pfldb#faq","mainEntity":[{"@type":"Question","name":"What is the difference between AI pentesting and pentesting an AI system?","acceptedAnswer":{"@type":"Answer","text":"They are two different jobs that share a confusing phrase. AI pentesting means an autonomous agent runs a penetration test against your own web apps and APIs, reproducing attacker behavior like reconnaissance and request chaining. Pentesting an AI system instead means assessing the model itself, its prompts, training data, retrieval layer, and the tools it can call, which is also called AI red teaming. Teams sorting out which of these two testing jobs they actually need can track the distinction through coverage like this on daily.dev."}},{"@type":"Question","name":"Does adding a second test identity actually find more vulnerabilities in an AI pentest?","acceptedAnswer":{"@type":"Answer","text":"Yes, adding a second account to an AI pentesting agent typically surfaces 30 to 50 percent more issues, because an agent testing with only one identity has nothing to compare against and cannot detect broken object-level authorization. A benchmark that held the model constant but changed only the harness found four times as many real findings at a zero percent false-positive rate. Developers comparing pentesting tool setups can follow benchmarks like this one on daily.dev before picking a vendor."}},{"@type":"Question","name":"Can AI agents find vulnerabilities that traditional vulnerability scanners miss?","acceptedAnswer":{"@type":"Answer","text":"Yes, AI agents can chain multi-step, identity-dependent flaws that scanners structurally cannot test, such as whether one user can reach another user's data, whether a business process can be completed out of order, or whether a bug chains into remote code execution. Google's Big Sleep agent found an exploitable stack buffer underflow in SQLite, described by Project Zero as the first public case of an AI agent finding a previously unknown exploitable memory-safety issue in widely used real-world software. Engineers deciding whether agentic tools add real coverage beyond scanners can follow these findings on daily.dev."}}]} ```