AI-Assisted Vulnerabilities Show Low Exploitation Rates in 2026

AI-Assisted Vulnerabilities Show Low Exploitation Rates in 2026

First seen 28 Jul 2026, 21:24 UTC Theregisterwww.vulncheck.com 79% similarity 42.9

Article Content

Browse articles
ThreatCluster

In the first half of 2026, VulnCheck reported a significant increase in CVEs, with 495 known exploited vulnerabilities (KEVs) identified. Despite the rise in AI-assisted vulnerability discovery, only 1.3% of AI-found vulnerabilities have been exploited, similar to the overall exploitation rate. The report indicates that while AI tools like Anthropic's Project Glasswing have identified numerous potential flaws, the actual weaponization of these vulnerabilities remains low. The median time from CVE publication to KEV status decreased from 120 days to 80 days, suggesting faster recognition of threats. CISA has issued guidance recommending rapid remediation for vulnerabilities with evidence of exploitation. The overall KEV-to-CVE ratio has declined, indicating that while more vulnerabilities are being disclosed, fewer are being exploited. This trend raises questions about the effectiveness of AI in enhancing real-world attack capabilities.

Key Points: • Only 1.3% of AI-assisted vulnerabilities have been exploited, similar to overall rates. • 495 known exploited vulnerabilities were identified in the first half of 2026. • CISA recommends rapid remediation for vulnerabilities with evidence of exploitation.

ThreatCluster AI How this analysis works

Timeline

2026-01-23
CVE-2026-0769 published
A vulnerability was disclosed, but no exploitation details are provided.
www.vulncheck.com
2026-02-20
CVE-2026-26980 published
A new vulnerability was published, with a PoC released on 2026-03-30.
www.vulncheck.com
2026-03-27
CVE-2026-5027 published
A vulnerability was disclosed, with a PoC available shortly after on 2026-04-02.
www.vulncheck.com
Recent
CISA issues guidance
CISA recommends prioritizing remediation for vulnerabilities with evidence of exploitation.
www.vulncheck.com
Recent
AI vulnerability findings analyzed
VulnCheck analyzed 1,061 AI-assisted discoveries, confirming only 14 exploited in the wild.
Theregister

Community

Browse all →