Benchmark RadarAI BENCHMARK PROFILE
SecDrift
SecDrift is a framework measuring sector-conditioned security drift in AI-generated code, evaluating vulnerability rates from LLMs when prompted with industry contexts versus neutral baselines across CISA sectors and CWE categories.
- Released
- 2026-07-28
- Readiness
- Paper only
- Primary field
- Cybersecurity
Why it matters
It addresses whether domain-specific prompting affects code security, with findings that model choice matters more than prompt framing. The framework could guide deployment decisions, but its standalone benchmark status is unclear.
Motivation
LLMs are increasingly used for code generation in critical infrastructure, yet the security effect of domain-specific prompting is understudied.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.