Policy pack · Compliance
EU AI Act, high risk policy pack
Policy set for EU AI Act compliance targeting HIGH risk AI systems under Annex III. Implements requirements from Articles 6-15 covering:
What's in it
12 policies, 4 of them blocking.
Critical policies block the step before it runs. High, medium and low ones let it run and open an incident.
| Policy | Checked | Severity | Rule |
|---|---|---|---|
| Agent must specify risk classification | At registration | critical | field_not_empty |
| Risk classification must be valid | At registration | critical | field_in_list |
| Agent must have a documented purpose | At registration | high | field_not_empty |
| Agent must have an identifying name | At registration | high | field_not_empty |
| Agent must have a designated owner | At registration | high | field_not_empty |
| HIGH risk agents must have detailed purpose documentation | At registration | high | field_matches_regex |
| HIGH risk agents must have a designated maintainer | At registration | medium | field_not_empty |
| HIGH risk - Mutating resource operations require human approval | Before every step | critical | step_requires_gate |
| HIGH risk - Destructive operations require human approval | Before every step | critical | step_requires_gate |
| HIGH risk - Data quality gate required before model calls | Before every step | high | step_requires_predecessor |
| Limit resource calls per task for robustness | Before every step | high | execution_max_steps |
| Limit model calls per task for robustness | Before every step | high | execution_max_steps |
Version 1.0.2. Source: manifests/compliance/eu-ai-act-high-risk.yaml.
Assign it
One command, no redeploy.
Assign it from the dashboard or the CLI. Running agents pick up an assignment within the policy refresh window, five minutes by default.
- Policies
- 12, of which 5 are checked before every step.
- Version
- 1.0.2
$ kyvvu list-manifests$ kyvvu assign-manifest --agent-id <agent> --repo-id <repo> \ --manifest manifests/compliance/eu-ai-act-high-risk.yamlNot legal advice. This pack is Kyvvu's interpretation of the EU AI Act. It isn't reviewed or endorsed by a regulator, and using it doesn't on its own make you compliant: it gives you technical controls, and a record that they ran, to support your compliance work.
About this pack
- Risk classification validation (Article 6) - Technical documentation (Articles 11, 13, Annex IV) - Human oversight gates on mutating and external operations (Article 14) - Data governance checkpoints (Article 10) - Robustness controls via execution limits (Article 15)
Policies without a risk_classification field apply to all agents. Policies with risk_classification: high apply only to HIGH risk agents.
Runtime policies in this pack
- HIGH risk - Mutating resource operations require human approval (critical)
- HIGH risk - Destructive operations require human approval (critical)
- HIGH risk - Data quality gate required before model calls (high)
- Limit resource calls per task for robustness (high)
- Limit model calls per task for robustness (high)
From Kyvvu's open policy library at github.com/Kyvvu/manifests, Apache 2.0.
Next
Run the whole pack in your browser.
In your browser
Try it in the playground
The policy is loaded. Add the steps your agent takes and watch the engine decide, step by step.
On your machine
Five minutes, no account
kyvvu try runs a demo agent against the bundled baseline. Then assign this pack from the dashboard or the CLI.
For security teams
See it on your agents
A 30-minute demo on the agents you actually run, with the policies that fit them.