The Exploit Was the Bottleneck
Summary
- A single operator using cheap, open-weight AI agents compromised 395 organizations in 48 countries within hours by exploiting two bugs disclosed just days earlier, with half the victims being schools.
- Critical PHP bugs leaked 7 weeks after a previous incident, leading to exploit attempts on Amazon decoy servers 4 days before official disclosure & working shells appearing within 3 days.
- The time from vulnerability disclosure to attack has collapsed from 63 days in 2018-2019 to an average of one day before disclosure in 2024 & seven days before in 2025.
- AI has reduced the cost & time required to create exploits from weeks or months to minutes or hours, allowing known bugs to be weaponized before most organizations can read advisories.
- Patching speeds have not improved; the median time to fully fix a known-exploited vulnerability is 43 days & only 26% of such vulnerabilities were fixed within a year in 2025.
- Grid & industrial control systems remain vulnerable because they cannot update at machine speed & some lack patches entirely, though physical architecture limits the potential for total collapse.
- AI is shifting the threat landscape toward the "long tail" of small targets like schools, hospitals, & water utilities that were previously too small to target due to high attack costs.
- Ransomware payments fell to ~$820 million in 2025 while leak-site claims rose, indicating attackers are targeting smaller victims who cannot pay & relying on destructive attacks instead.
- Recent large-scale incidents included the PaperCut campaign, supermarket refrigeration failures across multiple European countries, & a $6.34 billion theft from Binance via compromised engineer laptops.
- Experts warn that unrestricted AI models will soon automate the full attack chain without human intervention, potentially turning local disruptions into regional emergencies.
- Key indicators to watch include the total number of exploited vulnerabilities, the ratio of attacks occurring at disclosure, & the speed of patching relative to exploitation.
- Defenders must shift focus from pure patching to isolation, segmentation, & network-level protection for systems where updates are slow or physically impossible to apply.
What 23 Years of Data Say About AI and the Patch Gap
- Keywords
- Growth, Infra