Compliance dossier — training data (AI Act, Article 10)

Automatically generated skeleton from the dataset analysis. The "to be completed by the expert" blocks require human review (provenance, compliance judgment, mitigation measures).

1. Identity & purpose

System / model: _____
Dossier version: _____
Analysis date: _____
Dataset: src
Intended purpose: _____

_To be completed by the expert — intended purpose and what the data is meant to represent_

2. Provenance & legal basis

Origin of each source, licences/contracts, and — for personal data — initial purpose and GDPR legal basis.

⚠️ Potential personal data detected: telephone — a GDPR legal basis is required for these columns.

_To be completed by the expert — provenance per source (separate flow §16), licences, GDPR legal basis_

3. Composition _(filled automatically)_

Volume: 1000 rows, 21 columns.
Variables:

Column	Type
`status`	VARCHAR
`duration`	BIGINT
`credit_history`	VARCHAR
`purpose`	VARCHAR
`amount`	BIGINT
`savings`	VARCHAR
`employment_duration`	VARCHAR
`installment_rate`	BIGINT
`personal_status_sex`	VARCHAR
`other_debtors`	VARCHAR
`present_residence`	BIGINT
`property`	VARCHAR
`age`	BIGINT
`other_installment_plans`	VARCHAR
`housing`	VARCHAR
`number_credits`	BIGINT
`job`	VARCHAR
`people_liable`	BIGINT
`telephone`	BOOLEAN
`foreign_worker`	BOOLEAN
`credit_risk`	BIGINT

Declared sensitive attributes: status, credit_history, purpose, savings, employment_duration, personal_status_sex, other_debtors, property, age, other_installment_plans, housing, job, foreign_worker.

Detected PII columns: telephone.

_To be completed by the expert — geographic / contextual / behavioural scope_

4. Preparation _(automatic observations)_

_To be completed by the expert — transformation log: collection, cleaning, labelling, enrichment, aggregation_

5. Quality _(filled automatically)_

Accuracy / outliers: duration (1).

_To be completed by the expert — representativeness and relevance to the purpose_

6. Bias

Automatic analysis

Characteristic	Measure	Target	Verdict
Completeness	100.0 %	≥ 95.0 %	compliant
Uniqueness	100.0 %	≥ 99.0 %	compliant
Validity	99.7 %	≥ 98.0 %	compliant

status — « credit_risk »: disparate impact (ratio) 1.74 · statistical parity (gap) 37.6 % ⚠️

Group	Count	« credit_risk » rate
`no checking account`	394	88.3 %
`... >= 200 DM / salary for at least 1 year`	63	77.8 %
`0 <= ... < 200 DM`	269	61.0 %
`... < 100 DM`	274	50.7 %

credit_history — « credit_risk »: disparate impact (ratio) 2.21 · statistical parity (gap) 45.4 % ⚠️

Group	Count	« credit_risk » rate
`critical account/other credits existing`	293	82.9 %
`delay in paying off in the past`	88	68.2 %
`existing credits paid back duly till now`	530	68.1 %
`all credits at this bank paid back duly`	49	42.9 %
`no credits taken/all credits paid back duly`	40	37.5 %

purpose — « credit_risk »: disparate impact (ratio) 1.59 · statistical parity (gap) 32.9 % ⚠️

_⚠︎ Non-robust disparity: smallest group n=9 — not statistically significant or driven by a small sample. Do not read as an established bias._

Group	Count	« credit_risk » rate
`business`	9	88.9 %
`car (used)`	103	83.5 %
`domestic appliances`	280	77.9 %
`radio/television`	181	68.0 %
`repairs`	12	66.7 %
`others`	97	64.9 %
`education`	22	63.6 %
`car (new)`	234	62.0 %
`furniture/equipment`	12	58.3 %
`retraining`	50	56.0 %

savings — « credit_risk »: disparate impact (ratio) 1.37 · statistical parity (gap) 23.5 % ⚠️

Group	Count	« credit_risk » rate
`... >= 1000 DM`	48	87.5 %
`500 <= ... < 1000 DM`	63	82.5 %
`unknown/no savings account`	183	82.5 %
`100 <= ... < 500 DM`	103	67.0 %
`... < 100 DM`	603	64.0 %

employment_duration — « credit_risk »: disparate impact (ratio) 1.31 · statistical parity (gap) 18.3 % ⚠️

Group	Count	« credit_risk » rate
`4 <= ... < 7 years`	174	77.6 %
`... >= 7 years`	253	74.7 %
`1 <= ... < 4 years`	339	69.3 %
`unemployed`	62	62.9 %
`... < 1 year`	172	59.3 %

personal_status_sex — « credit_risk »: disparate impact (ratio) 1.22 · statistical parity (gap) 13.4 % ⚠️

Group	Count	« credit_risk » rate
`male : single`	548	73.4 %
`male : married/widowed`	92	72.8 %
`female : divorced/separated/married`	310	64.8 %
`male : divorced/separated`	50	60.0 %

other_debtors — « credit_risk »: disparate impact (ratio) 1.44 · statistical parity (gap) 24.7 % ⚠️

Group	Count	« credit_risk » rate
`guarantor`	52	80.8 %
`none`	907	70.0 %
`co-applicant`	41	56.1 %

property — « credit_risk »: disparate impact (ratio) 1.39 · statistical parity (gap) 22.2 % ⚠️

Group	Count	« credit_risk » rate
`real estate`	282	78.7 %
`building society savings agreement/life insurance`	232	69.4 %
`car or other`	332	69.3 %
`unknown/no property`	154	56.5 %

other_installment_plans — « credit_risk »: disparate impact (ratio) 1.23 · statistical parity (gap) 13.5 % ⚠️

Group	Count	« credit_risk » rate
`none`	814	72.5 %
`stores`	47	59.6 %
`bank`	139	59.0 %

housing — « credit_risk »: disparate impact (ratio) 1.25 · statistical parity (gap) 14.6 % ⚠️

Group	Count	« credit_risk » rate
`own`	713	73.9 %
`rent`	179	60.9 %
`for free`	108	59.3 %

job — « credit_risk »: disparate impact (ratio) 1.10 · statistical parity (gap) 6.5 % ⚠️

_⚠︎ Non-robust disparity: smallest group n=148 — not statistically significant or driven by a small sample. Do not read as an established bias._

Group	Count	« credit_risk » rate
`unskilled - resident`	200	72.0 %
`skilled employee/official`	630	70.5 %
`unemployed/unskilled - non-resident`	22	68.2 %
`management/self-employed/highly qualified employee/officer`	148	65.5 %

foreign_worker — « credit_risk »: disparate impact (ratio) 1.29 · statistical parity (gap) 19.9 % ⚠️

Group	Count	« credit_risk » rate
`false`	37	89.2 %
`true`	963	69.3 %

Judgment & measures

_To be completed by the expert — impact on health/safety/fundamental rights and detection/prevention/mitigation measures_

7. Gaps & limitations

Detected points to examine as potential gaps:

[low] outliers — duration
[low] outliers — amount
[high] pii — telephone
[medium] bias — status
[high] bias — credit_history
[low] bias — purpose

_To be completed by the expert — identified and addressed gaps; out-of-scope uses_

8. Governance & traceability

Analysis tool version: 0.4.1

Dataset fingerprint (schema + volume): ca33a8c44d3690a0

_To be completed by the expert — responsibilities, audit log, versioning, maintenance/updates_