SFS E: AI-Oriented Ethics (Training and Implementation)
Assess the ethical and accountable practices used to source data, develop, train, test, and prepare AI systems for deployment.
Scope and boundaries
This set covers the provenance and permitted use of training data, documentation of model choices, evaluation of bias and fairness, privacy considerations, and responsibility for approving deployment. Questions should state the intended use, relevant populations, evidence, and acceptance criterion. Ethical findings must remain distinguishable from security controls such as access permissions and protection of training inputs. General engineering practices belong to SFS 7 and SFS 8; behavior and oversight after deployment belong to SFS F.
The sixteen questions
Each SF position is one question. Both positions run from 0 to F.
| SF | Question page | Authoring record |
|---|---|---|
| 0 | SF 0 | 00E0 |
| 1 | SF 1 | 00E1 |
| 2 | SF 2 | 00E2 |
| 3 | SF 3 | 00E3 |
| 4 | SF 4 | 00E4 |
| 5 | SF 5 | 00E5 |
| 6 | SF 6 | 00E6 |
| 7 | SF 7 | 00E7 |
| 8 | SF 8 | 00E8 |
| 9 | SF 9 | 00E9 |
| A | SF A | 00EA |
| B | SF B | 00EB |
| C | SF C | 00EC |
| D | SF D | 00ED |
| E | SF E | 00EE |
| F | SF F | 00EF |
From questions to a profile
Only reviewed, accepted 0/1 assessments can populate the sixteen-bit set profile. SF 0 is the most significant bit and SF F the least significant bit. The completed result uses four hexadecimal digits, including leading zeros. A catalog definition does not establish a completed assessment profile.
Read assessment and validation, then encoding. A set profile preserves individual answers; it is not a summed security score.