Public-sector AI deployment policy
DSarchitek Inc. writes AI policy for public bodies around one rule: AI tools do not get direct access to secured personal information.
Why the rule comes first
AI tools are easy to connect to data and hard to audit once connected. A policy that starts from a list of approved tools goes out of date within months. A policy that starts from what the tools may touch stays useful as the tools change.
The rule then shapes the rest: how AI is implemented across government data sources, where synthetic data stands in for real records, how vendor claims are tested, and who oversees how data is used.
What the practice does
- AI policy frameworks for ministries and public bodies.
- AI governance frameworks inside wider data strategies.
- Synthetic data approaches that let AI tools be built and tested without secured personal information.
- Ways to test vendor claims before a contract is signed.
- Policy written with the Treasury Board Directive on Automated Decision-Making and the EU AI Act in view.
Work behind it
On a Canadian provincial registries program, the practice wrote the AI policy framework that keeps AI tools away from secured personal information by using synthetic data, and the AI governance framework within the ten-year data modernisation strategy. At Export Development Canada, the principal advised the cyber risk and data governance teams on anonymisation and data masking.
The practice also licenses The Deontic Architecture®, a reference architecture that lets a person contest results an AI model derives from their records.