Cybersecurity Statistics

Cybersecurity statistics / Vulnerability Research

A year ago, 55% of AI models failed basic vulnerability research and 93% failed exploit development tasks

Publisher	Forescout
Report	AI Security Testing: Agents Leap from Assistants to Autonomous Hackers
Published	14 April 2026
Topics	Vulnerability Research, Exploit Development, AI Security, AI Models

Published by Forescout in AI Security Testing: Agents Leap from Assistants to Autonomous Hackers , 14 April 2026. The figure is taken from the report as published; the full methodology is in the source.

View the original report

Compare this across sources

12 sources answer what share of organisations have an ai governance policy in place. They report between 18% and 44%, with a median of 36.5%.

See all 12 sources

Related statistics

Every model produced at least one false-positive run by hallucinating vulnerable paths in the OpenNDS real-world task

Forescout, 27/05/2026

All tested AI models now complete vulnerability research tasks, and 50% generate working exploits autonomously

Forescout, 27/05/2026

Claude Opus 4.6 identified exploitable RCE vulnerabilities in 3 of 5 runs.

Forescout, 27/05/2026

AI-assisted exploit development compressed the average time from vulnerability disclosure to a working exploit from 125.3 days in January 2025 to 0.5 days by April 2026.

Cogent Security, 31/05/2026

62% of enterprises are concerned about training data breaches in AI workflows.

Perforce Software, 25/07/2026

Only 7.6% of surveyed OT and ICS cybersecurity professionals are very confident in controls that protect AI tools and models because those controls have been tested.

Nozomi Networks, 25/07/2026

Get the newsletter

Weekly cybersecurity statistics by email.