Only 4% of security leaders believe their current testing frequency gives them sufficient coverage.
Deployment cadence outpaces pentesting schedules
Software is moving faster: 68% of organizations deploy code to production at least daily, and 37% deploy multiple times a day. Pentesting has not kept pace. Nearly half of respondents (47%) go more than a month between full pentests on their most frequently tested web applications and APIs. The mismatch matters: a vulnerability introduced in one release can sit untested for weeks or months, and most security leaders feel the gap. Four in ten (42%) say they simply accept the risk and ship code without additional security checks between pentests.
AI pentesting adoption is rapid and uneven
AI-driven pentesting has moved from experimental to mainstream quickly. Intruder’s survey of 500 security and engineering leaders across the US and UK found 55% of organizations already use AI pentesting across their environment, and a further 36% have tried it but not yet scaled it. Fewer than 1% have no plans to try it. Larger organizations — those with 5,000 to 10,000 employees — are furthest along, with 71% using AI pentesting widely; midmarket organizations have tried it at a similar rate but scaled adoption less (54%).
Attitudes toward traditional testing have shifted decisively: 90% agree that “the traditional annual pentest is dead” and that continuous AI pentesting is required to keep pace. Only 6% still rely on a single annual pentest. Looking three to five years ahead, 54% expect AI to handle the bulk of testing, one in six expect manual pentests to remain the primary method, and 28% expect to keep manual pentests mainly to satisfy auditors.

This site is the portfolio.
OSINTSights runs on Cloudflare Workers, D1, R2, and Vectorize, with an AI pipeline on Hetzner ARM. Nubivance designed, built, and operates it. We do the same for clients.
See what we buildData privacy and production safety top the barrier list
Quality concerns are real but not dominant: only a quarter of respondents name AI testing quality as a top barrier to adoption. The more pressing worries are operational and legal. Forty percent point to data privacy as a top barrier, 36% cite safety on production systems, and 28% worry whether auditors and regulators will accept AI pentest reports.
Trust hinges on visibility and reproducibility. Asked what would most damage their trust in AI pentesting, nearly half (46%) pointed to a lack of visibility into what was tested (25%) or a lack of reproducible proof of exploitability (21%). The survey also found role-based differences: DevSecOps leads — the people handling tool output day-to-day — are the most skeptical, with only 5% saying nothing would damage their trust, compared with 19% of CTOs.
Operational friction: triage, false positives and slow retesting
The mechanics of vulnerability validation drive much of the appetite for change. More than half of teams (56%) spend six or more hours a week validating findings from their security tooling; one in five spends 11 hours or more. Working out whether a finding is real is the biggest remediation bottleneck, cited by 44% of leaders.
Confirming fixes is slow as well. Fewer than two in five organizations independently retest a fix in production as soon as it ships; nearly half take up to two weeks. That gap leaves a long window where a reported vulnerability may or may not actually be closed. Intruder frames AI-enabled validation as specifically designed to reduce that wasted effort by confirming exploitability before issues are reported and by retesting fixes immediately after deployment.
What this means for DevSecOps leads, CTOs, and security teams
- DevSecOps leads: Expect heightened demands for transparency and reproducible proof-of-concept. The survey shows they are the most skeptical role; visibility into what was tested and concrete exploitability evidence are conditions for trust.
- CTOs: Compared with operational teams, CTOs are more likely to tolerate some uncertainty — 19% say nothing would damage their trust — but they still face the strategic choice of whether to scale AI testing across fast deployment pipelines.
- Security teams: The fast adoption of AI pentesting offers a way to align testing cadence with deployment cadence and to cut the hours spent triaging false positives, but teams must resolve data-handling and production-safety issues before granting tools broad access.
Intruder’s findings sketch a simple, consequential trade: AI pentesting can bring pentest depth to every release, but the chief barriers are not model quality alone — they are how the tools operate inside sensitive environments and whether their results are visible and reproducible. Practical early steps the survey endorses are straightforward: start in a staging environment that mirrors production, run AI pentesting alongside a manual pentest and compare results, require proof-of-concept code and a record of what was tested, and check how code and data will be handled before granting access. The wider adoption of AI pentesting now depends less on the promise of better coverage and more on solving those operational trust and safety questions — including whether auditors and regulators will accept the outputs.




