AI models and infrastructure

AI safety testing

Tools for detecting unsafe outputs, jailbreaks, prompt injection, and other model risks

Ai safety testing is growing in line with its peers; its strongest place is open models at 1.1× the typical product. The fastest product is prompt-guard-68m, up 17×. 82 products measured, as of Oct 10, 2026.

How we measure

Where it grows against its peers

  • Open models45 products1.10×

The line marks 1×: the typical product in that place.

Fastest products

  1. 1prompt-guard-68mOpen models17×
  2. 2AI Agent Security Scanner for IDEsOpen-source editor extensions1.9×
  3. 3VibeGuard AICodingOpen-source editor extensions1.7×
  4. 4ncii-guard-v02Open models1.6×
  5. 5gws-modelarmor-create-templateAgent skills1.2×
  6. 6gws-modelarmor-sanitize-responseAgent skills1.2×

AI safety testing: quick answers

Not faster than its peers right now; the closest is open models at 1.1× the typical product.

More in AI models and infrastructure

Model inference services11.3×Video generation models5.6×Image generation models5.1×Speech and audio models1.9×Language model apis1.7×Roleplay and storytelling models1.6×AI retrieval systems1.4×Prompt management1.3×