November 20, 2026

Software flaws now outpace fixes 18 to 1 as AI shrinks the window to exploit them

Black-Duck-AI-readiness-report

Organisations are discovering software security flaws 18 times faster than they are fixing them, according to new research published today by application security company Black Duck. Meanwhile, the time attackers need to exploit a newly disclosed flaw has fallen from years to less than a day.

Black Duck’s 2027 AI Readiness Report analysed 65.5 million vulnerabilities found across around 13,100 applications at 533 organisations between June 2025 and May 2026. It found that by the end of the study, only 10.5% of those flaws had been fixed.

The gap between finding flaws and fixing them widened over the year. In June 2025, organisations found 11 new flaws for every one they closed. By May 2026 the ratio had reached 18 to 1, after peaking at 22 to 1 in December 2025. In April 2026 alone, organisations added 10.4 million unresolved flaws to their backlogs.

Attackers moving faster

The report says the growing backlog matters because attackers are moving faster than ever. In 2018, the average time between a vulnerability being disclosed and first being exploited was 2.6 years. It is now less than a day, according to the Zero Day Clock.

The report links part of that acceleration to AI. It cites lab tests in which Anthropic’s Claude Mythos Preview model produced 181 working exploits against the Firefox browser, compared with two from its predecessor, and an AI-driven attack documented by Sysdig in February 2026 that reached admin-level access in eight minutes.

Fixes take far longer. The fastest-fixed flaws in the study, found in companies’ own code and in open source components, took a median of 28 and 39 days to close. Flaws found through manual testing took a median of 232 days, and those found through dynamic testing took 316 days.

Serious flaws left until later

The research also suggests that the most dangerous flaws are not always fixed first. Among open source components, critical vulnerabilities took a median of 58 days to close, against 22 days for high-severity ones. Black Duck said this suggests teams may be clearing the easiest issues first rather than the most dangerous.

Open source and third-party code accounted for 52% of all critical and high-severity vulnerabilities in the study, yet only 15% of open source vulnerabilities had been closed.

“Finding vulnerabilities is no longer the hard part,” said Greg Hughes, CEO of Black Duck. “Boards and regulators will judge organisations on time to fix, not on how many vulnerabilities they found.”

Regulatory pressure

The report highlights growing regulatory pressure to fix flaws quickly. The PCI DSS 4.0.1 payment card standard requires critical security patches within one month of release, but nearly half of the critical open source flaws that were fixed in the study took longer than that. The EU Cyber Resilience Act, which affects UK firms selling connected products into the EU, introduced reporting obligations in September 2026, including an early warning on actively exploited vulnerabilities within 24 hours.

Scanning frequency linked to faster fixes

Some organisations are keeping pace better than others. Those that scanned their open source components 10 to 29 times a month fixed flaws in an average of 30 days, compared with 66 days for those scanning less than once a month. Organisations with more than 10,000 employees fixed their critical and high-severity flaws in a median of 12 days, faster than their 19-day median for all flaws.

Black Duck recommends that organisations combine consistent, automated security testing with AI-assisted triage to decide which flaws to fix first, and AI-generated fixes that developers review and approve.

The research is based on aggregated, anonymised data from Black Duck customers. Because the sample is made up of organisations that invest in commercial application security tools, the company notes that they may be more mature than the wider market.