Scanning 3,911 AI-built app repositories: risk rises with code volume
Understand this story
Scanner vendor Path to Ship analyzed 3,911 public repositories produced by nine AI build tools—Bolt, Lovable, Cursor, v0, Replit, Windsurf, Base44, Tempo, and Firebase Studio—and found that 66.3% have no stars, forks, or issues, 68.1% have not been pushed to in the past 90 days, and most vulnerabilities fall in dormant apps that no one uses. The study also revised an earlier attribution figure of 92.8% down to 78.3%. Path to Ship says the only robust finding is that the larger an app's codebase, the higher the share containing at least one severe issue, rising monotonically: 2.9% for under a thousand lines, reaching 59.2% above forty thousand lines; it suggests apps of a few hundred lines with no users need not worry, while those above ten thousand lines with users carry a baseline risk of about 30% that rises as features are added. The study covers only public repositories and does not include private ones; Path to Ship says that given the dormancy, private repositories may be the less problematic half, and recommends reading absolute percentages as an upper bound and the trend of rising risk with code volume as the conclusion until the sample is completed.
AI overview based on source reports · Updated 14 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Hacker News AIWhen AI-built apps get risky: data from 3,911 repos
An analysis of 3,911 public repositories across nine AI builders found that 66.3% have no stars, forks or open issues and 68.1% have not been pushed in 90 days, so most flaws sit in dormant apps. The share of apps with at least one critical finding rises monotonically with size, from 2.9% under 1,000 lines to 59.2% over 40,000, a gradient that held across all nine builders and both scanner builds. The authors caution that the scanner over-flags, that public repos are not a representative sample of deployed apps, and that the data measures code rather than breaches.
Activity for this story
- Current comparable activity
- 7
- Peak comparable activity
- 10Nov 20 06:00
- Change in the past 24 hours
- –
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.