Skip to content
Quality Isn’t The Main Worry About AI Pentesting — Data Privacy and Production Safety Are

Quality Isn’t The Main Worry About AI Pentesting — Data Privacy and Production Safety Are

AI is speeding up how software is built, which means more code, more changes, and more chances for vulnerabilities to reach production. It’s speeding up how attackers find and exploit them too, with time-to-exploit now down to a single day. Security testing has to keep up with both.

AI pentesting has emerged as a way to run in-depth tests whenever teams need them, instead of waiting for the next scheduled pentest. To see how far adoption has come and what security teams need from it, Intruder surveyed 500 security and engineering leaders across the US and UK. Read on for the top findings.

Only 4% of security leaders believe their current testing frequency gives them sufficient coverage

68% of organizations deploy code to production at least daily, and 37% deploy multiple times a day. Understandably, pentesting runs on a much slower cycle. Nearly half (47%) go more than a month between full pentests on their web applications and APIs. Those figures are for each organization’s most frequently tested application, so the rest of their estate is tested less often still.

Running a human-led pentest as often as code ships has never been realistic, largely because of the cost. But it means a vulnerability introduced in one release can go weeks or months without a pentest, and only 4% of security leaders believe their current testing frequency gives them sufficient coverage.

42% say they accept the risk and ship code without additional security checks between pentests

Most teams cover the gap between pentests with a mix of approaches, but 42% accept the risk and ship code without additional security checks in between. Automated scanning in the build pipeline, more frequent manual pentests, internal AppSec teams, and AI-driven pentesting are each used by around half. 

Concern about the gap is widespread. 88% of leaders running scheduled pentests are worried about unmapped risk introduced between tests, and they split almost evenly on how serious it is. Half call it one of their largest unmanaged cyber risks, while the other half describe it as a known gap they’re covering with basic perimeter controls.

Fewer than 1% have no plans to try AI pentesting

55% of organizations already use AI pentesting across their environment, a further 36% have tried it but not yet adopted it at scale, and fewer than 1% have no plans to try it. For a category that has only emerged in the past two years, that’s fast uptake.

The largest organizations surveyed (5,000 to 10,000 employees) are furthest along, with 71% using AI pentesting across their environment. Midmarket organizations have tried it at a similar rate but scaled it less so far, at 54%. 

Only a quarter see AI testing quality as a top concern

Security leaders need confidence that AI pentesting is safe to run on sensitive systems, and a way to check its work. The quality of AI pentesting isn’t the main concern, with only a quarter naming it as one of their biggest barriers to adoption. The top barriers relate to how AI tools operate inside sensitive environments: data privacy (40%), safety on production systems (36%), and whether auditors and regulators will accept AI pentest reports (28%).

Asked what would most damage their trust in AI pentesting, nearly half (46%) point to a lack of visibility into what was tested (25%) or a lack of reproducible proof of exploitability (21%).

DevSecOps leads, who work most directly with the findings security tools produce, are the most skeptical — only 5% say nothing would damage their trust, compared with 19% of CTOs.

56% of teams spend six or more hours a week triaging false positives

More than half of teams (56%) spend six hours or more a week validating findings from their security tooling. Working out whether a finding is real is the biggest remediation bottleneck, named by 44% of leaders, and one in five teams spend 11 hours or more a week on it, more than a full working day.

Confirming fixes takes time too. Fewer than two in five organizations independently retest a fix in production as soon as it ships, and nearly half take up to two weeks. Until then, there’s no way to know the vulnerability has actually been closed.

This is the work AI-enabled validation is designed to cut, by confirming findings are exploitable before they reach the team and retesting a fix as soon as it ships.

90% agree the traditional annual pentest is dead

Nine in ten security and engineering leaders agree that the traditional annual pentest is dead and that continuous AI pentesting is now required to keep pace. For most, it already is. Only 6% still rely on a single annual pentest.

Looking three to five years ahead, most (54%) expect AI to handle the bulk of testing, and only one in six expect manual pentests to remain a primary method of security validation. 28% expect to keep manual pentests mainly to satisfy auditors, regardless of their technical value.

What this means for security teams

Until now, teams have relied on two kinds of testing, each with a trade-off. Scanning and DAST run often but are generic by design, so they miss flaws specific to how an application works. A pentest goes deep, but it’s a snapshot of how the application looked weeks or months ago. AI pentesting brings the depth of a pentest, on demand.

That changes what’s possible across the testing cycle. Releases can be tested as they ship, instead of waiting for the next scheduled pentest. With findings validated before they’re reported, teams spend their time on issues that are real, and fixes can be retested straight away.

For teams looking to adopt it with confidence, a few steps help. Start in a staging environment that mirrors production. Run an AI pentest alongside your next manual pentest and compare the results. Look for findings that come with proof-of-concept code and a record of what was tested. And check how your code and data will be handled before you give it access.

With AI speeding up how software is built and how it’s attacked, AI pentesting makes it possible to keep pace with both. Get the full report or learn more about Intruder’s AI pentesting.

Found this article interesting? This article is a contributed piece from one of our valued partners. Follow us on Google News, Twitter and LinkedIn to read more exclusive content we post.



Source link