Updated

Which AI model writes the most security and privacy bugs?

As of Sep 27, 2026, Claude Sonnet 5 had the highest rate of security and privacy bugs, 81% above the all-model rate (1.81), statistically tied with GPT-5.6 Luna and Claude Opus 5. GPT-5.6 Sol had the lowest, 42% below average (0.58), statistically tied with GPT-5.5.

Models ranked by relative rate of security and privacy bugs, fewest first. Lower values are better.
RankModelRate
1= (statistically tied)GPT-5.6 Sol
0.58
2= (statistically tied)GPT-5.5
0.62
3= (statistically tied)Claude Fable 5.1
0.80
4= (statistically tied)GPT-6 Astra
0.82
5= (statistically tied)Claude Fable 5
0.88
6= (statistically tied)Claude Opus 4.8
1.08
7= (statistically tied)GPT-5.6 Luna
1.50
8= (statistically tied)Claude Opus 5
1.53
9= (statistically tied)Claude Sonnet 5
1.81

1.00 = all-model rate · lower is better · = statistically tied