Explains environment separation
Separate development, test and production (ISO/IEC 27002:2022 8.31), each with an owner and no production links.
Can AI do this?
AI does this partly Some of the tested AI models master this; others do not show it. 2 of 4 models in the current benchmark master it.
| Model | Tier | Probability of mastery | Status |
|---|---|---|---|
| gemini-4-argon | Flagship | 100% | Mastered |
| gemini-3.5-flash-lite | Low cost | 82% | Not shown |
| gpt-6-luna | Low cost | 100% | Mastered |
| claude-haiku-5-5 | Low cost | 97% | Not shown |
Mastered means a probability of at least 90% that the model has this knobit, estimated from its graded answers. “Not shown” means the answers do not settle it either way. How the models are tested.
What it looks like in practice
Expected result: Separate development, test and production (ISO/IEC 27002:2022 8.31), each with an owner and no production links.
How it is assessed: Answers a written question on the environment separation correctly
Related knobits
Needed for
- O-201 Recognises production personal data in non-production
- O-206 Recognises an environment with no owner
- O-207 Recognises non-production linked to production
All 15 knobits of Plan environments and capacity.
Source: Practice the stage 2 reference and review checklist rest on