AI writes half our code now. It still fails security tests 44% of the time.

Veracode's 2026 report finds AI code generation now produces compilable code nearly 100% of the time but still fails security tests 44% of the time, a rate unchanged over a year despite AI writing half of all code. Specialized coding models and larger model sizes do not improve security outcomes, while reasoning models perform slightly better. OpenAI's GPT-5.5 leads with a 68% security pass rate but has declined from last year.