New model's vuln stats look same as before: many found, few exploited in the wild

xeophon · x · 2026-09-21

xeophon sarcastically weighs in on the "most humans are good, so we'll be fine even with more cyber-capable models" camp.

The quoted post from GsInfosystems notes that the stats from the newer model (mythos) are similar to what came before: a large number of discovered vulnerabilities but only a very small percentage actually exploited in the wild — suggesting more capable models haven't materially shifted the cyber offense/defense landscape.

Original post →

More from Models

Models channel →