How we use AI

This document covers the GotAIPolicy.ca platform's own use of AI: summaries, comparisons, search, and analysis of user-uploaded policies. For how municipal policies in the registry are found, classified, and verified, see the methodology.

• Structured registry summaries (highlights, best practices, red flags, missing clauses).
• Side-by-side comparisons across multiple municipalities.
• Executive cohort briefs for Featured Cities.
• “Ask the registry” — Q&A grounded in registry evidence.
• “Review my policy” — structured analysis of a draft the user uploads or links.
• Search ranking and autocomplete suggestions.

We combine three Google Gemini models, all accessed through a managed AI gateway:

• google/gemini-2.5-flash — default engine for municipal policy summaries, "Review my policy", "Ask the registry", and side-by-side comparisons.
• google/gemini-2.5-pro — used for longer, more demanding cohort analyses (multi-city executive briefs).
• google/gemini-3-flash-preview — faster next-generation model used for AI-assisted FAQ search and autocomplete.

We do not train any models ourselves. Providers and model versions can change as quality improves. Material changes are logged in the changelog.

• Every registry row can carry a “human verified” flag when a reviewer has confirmed the cited evidence.
• All AI-generated text is labelled as such and treated as a starting point — never a decision. See the correction process.
• Founding Members have access to a priority verification queue and can request human review of specific entries.
• Users can challenge any status; submissions are reviewed against official public sources before a status changes.

• Registry features: only public, cited sources (council minutes, official policy URLs, municipal websites).
• “Review my policy”: the text you upload or link. That content stays confidential to your account and is not shared with other users.
• “Ask the registry”: your question plus cached public summaries.

• We do not directly train public AI models on user uploads, private drafts, direct messages, or forum content.
• We do not use AI to make automated decisions about people — no scoring of staff, residents, or vendors.
• We do not recommend or rank AI vendors.
• We do not fabricate quotes or sources: every citation links to a public, verifiable URL.
• We do not use AI to take punitive moderation actions on the forum without human review.

There is an important difference between training an AI model on user content and using anonymized, aggregated insights to improve a service.

Got AI Policy may analyze uploaded policies to generate a private review for the user. We may also use anonymized and aggregated analytical insights from many reviews to improve benchmarking, recommendation logic, prompt quality, safety checks, and user experience.

We do not use raw uploaded policies, private drafts, direct messages, or forum content to directly train public AI models. Aggregated insights are reviewed at a systems level and are not intended to identify users or organizations, reproduce original documents, or expose confidential content.

See the full Governance page →

AI-generated summaries can be wrong, outdated, or miss context. Before relying on any output — especially for a governance, legal, privacy, or procurement decision — you must verify it against the linked source.

This is the same rule we apply to the “Review my policy” feature: AI helps structure the analysis; it does not replace legal, privacy, or risk review.

The models we use are trained largely on global English data. We add Canadian municipal context through carefully written prompts and curated examples, but bilingual parity, Indigenous governance contexts, and small-municipality realities can still be under-represented.

If you notice bias or a blind spot, tell us — that's how the platform improves.

Uploads and user-generated content are retained only as long as needed to deliver the service and are not shared with model providers for training. You can delete your account and all of your data at any time from Settings → Danger zone. See the privacy policy for the full picture.

If an AI output is inaccurate, biased, or harmful, email support@gotaipolicy.ca or use the “Report issue” button on the affected page. We commit to investigate and correct.

This policy is informed by: Canada's Treasury Board Directive on Automated Decision-Making, ISED's Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems, and the OECD AI Principles. We do not claim any certification; we reference these frameworks to signal the standard we hold ourselves to.

Material changes to this policy will be logged in the changelog and the "Last updated" date above will be revised.