---
license: cc-by-4.0
language: [en]
task_categories: [text-classification]
tags: [eu-ai-act, ai-governance, legal, synthetic, devseis]
pretty_name: Devseis AI Act Classifier v6 data
size_categories: [1K<n<10K]
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train_v6.csv
  - split: validation
    path: data/val_v6.csv
  - split: test
    path: data/test_v6.csv
- config_name: scenarios
  data_files:
  - split: train
    path: data/scenarios_v6.csv
---

# Devseis AI Act Classifier — v6 training and evaluation data

*A Devseis dataset (`data_v6_systems`). Used to train [`Devseis/devseis-ai-act-classifier-v6`](https://huggingface.co/Devseis/devseis-ai-act-classifier-v6), which powers [Caveat](https://huggingface.co/spaces/Devseis/Caveat).*

This is the data the v6 classifier was trained and evaluated on. It is published so the numbers in the
model card and on the Caveat page can be checked. **It is synthetic, and it is not legal advice.**

## Files

| File | What it is | Rows | Scenarios |
|---|---|---|---|
| `data/train_v6.csv` | training set (`text`, `label`) | 938 | 242 |
| `data/val_v6.csv` | validation set, used only to choose settings | 162 | 54 |
| `data/test_v6.csv` | test set, used once per final configuration | 210 | 70 |
| `data/scenarios_v6.csv` | one row per distinct scenario (`text`, `label`, `legal_basis`) | 366 | 366 |
| `CHANGES_v3.md` … `CHANGES_v6.md` | what changed in each version and why | | |

Labels: `prohibited`, `high_risk`, `limited_risk`, `minimal_risk`. Counts by scenario:

| Split | high_risk | minimal_risk | prohibited | limited_risk |
|---|---|---|---|---|
| train | 96 | 76 | 43 | 27 |
| val | 19 | 17 | 10 | 8 |
| test | 27 | 21 | 12 | 10 |

## Loading

```python
from datasets import load_dataset
ds = load_dataset("Devseis/devseis-ai-act-classifier-v6-data")              # train / validation / test
scenarios = load_dataset("Devseis/devseis-ai-act-classifier-v6-data", "scenarios")
```

## How it was made

- **Synthetic, LLM-written scenarios.** Each is based on the text of Regulation (EU) 2024/1689
  (Art 5, Art 50, Annex I, Annex III), its recitals, and the European Commission's draft guidelines
  on classifying high-risk AI systems.
- **Template-expanded.** In validation and test, each scenario appears as 3 rows. In training, each scenario also
  has one extra row in a different voice (a question, a pitch, a tender request, …), added in v6 so the model sees
  more than one writing style.
- **Classic rows:** each scenario appears as 3 rows. The rows differ only in an opening phrase
  ("Our client wants to deploy…", "A startup is building…"). In about half the scenarios, one of the
  three rows has no opening phrase. The real sample size is the scenario count, not the row count.
- **Split by scenario.** All rows of a scenario sit in the same split. No scenario appears in more
  than one split. `test_v6` is the v5 test set plus 4 new scenarios (and the v5 test set is the v4 one plus 5), so
  results can be compared with earlier versions on the same rows.

## Known caveats

- Every scenario is synthetic, and the test sets are small (70 scenarios, 12 of them prohibited).
  Scores on this data are an upper bound on real-world performance.
- **Age and sex estimation from biometrics is labelled `high_risk`** (Annex III(1)(b)) as a consistent
  policy choice, because the law is unsettled on it.
- **The AI Omnibus scenarios** (non-consensual intimate imagery and AI-generated child sexual abuse
  material) are labelled `prohibited`, which is the law from **2 December 2026**. They are one-line,
  non-graphic descriptions of banned systems. Their `legal_basis` citations, Art. 5(1)(ba) and (bb), are
  verified against the Official Journal text of the AI Omnibus, Regulation (EU) 2026/1744, Art. 1(7).
  Under the new Art. 5(1a), a system is banned only where that output is its intended purpose, or a
  reasonably foreseeable outcome without adequate safeguards. All 9 scenarios describe systems built
  for that purpose.
- The `CHANGES_*.md` files were written for the build folders, so their relative paths (such as
  `../relabel_audit/`) point to files that are not published here.

## License

CC-BY-4.0.