Explains access design for classified data
Grants, masking and row-level security derived from data classification.
Can AI do this?
People still lead Few or none of the tested AI models showed they master this. 0 of 10 models master it.
| Model | Tier | Probability of mastery | Status |
|---|---|---|---|
| gemini-4-argon | Flagship | 62% | Not shown |
| gemini-3.1-pro-preview | Flagship | 60% | Not shown |
| gpt-6-astra | Flagship | 58% | Not shown |
| claude-fable-5-1 | Flagship | 63% | Not shown |
| gemini-3.8-flash | Mid tier | 60% | Not shown |
| gpt-6.1-sol | Mid tier | 61% | Not shown |
| claude-sonnet-5-5 | Mid tier | 60% | Not shown |
| gemini-3.5-flash-lite | Low cost | 61% | Not shown |
| gpt-6-luna | Low cost | 60% | Not shown |
| claude-haiku-5-5 | Low cost | 59% | 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: Grants, masking and row-level security derived from data classification
How it is assessed: Designs access for sample classified columns
Related knobits
Needed for
All 33 knobits of Generate and review DDL.
Source: Data-management practice used by the stage 8 reference and review checklist