Back to home
NorthNorth OS

Responsible AI Policy

Last Updated: 4 July 2026

North OS builds AI features intended to augment — never replace — human judgement. This policy describes the principles and governance we apply to the design, deployment and monitoring of AI in our platform, informed by UK ICO AI guidance and internationally recognised AI risk frameworks.

1. Fairness

  • We design AI features to serve all users equitably.
  • We avoid deploying AI in ways that could produce discriminatory outcomes against individuals or groups on the basis of protected characteristics.
  • Customers are contractually prohibited from using our AI for decisions where bias could cause legal or similarly significant harm without meaningful human review.

2. Bias Monitoring

  • We evaluate AI features for biased or skewed outputs before and after release.
  • User feedback and reported issues feed into regular bias reviews.
  • When material bias is identified, we adjust prompts, models, guardrails or feature design.

3. Human Oversight

  • AI features are designed with a human in the loop for consequential actions: drafts require review, automations are configurable and can be paused, and AI recommendations are advisory.
  • No North AI feature is designed to make final decisions with legal or similarly significant effects on individuals.

4. Explainability and Transparency

  • AI-powered features are identified as such in the product.
  • Where practical, AI outputs include the context or sources they drew from so users can verify them.
  • We document what our AI features do, their limitations, and appropriate use in our AI Usage Policy.

5. Model Governance

  • We maintain an inventory of the AI models and providers used in the platform.
  • New models and significant changes go through evaluation for quality, safety and privacy before deployment.
  • Provider agreements restrict training on customer data where such controls are available.
  • We monitor model performance in production and can roll back or switch models when issues arise.

6. Data Minimisation and Privacy

  • AI features process only the data needed for the requested task.
  • Customer Content is not used to train generally available models without consent.
  • AI processing is covered by our Privacy Policy and Data Processing Agreement.

7. Testing and Evaluation

  • AI features are tested against quality and safety criteria before release, including adversarial testing for prompt injection and misuse.
  • Beta AI features are labelled and gathered feedback informs improvements before general availability.
  • We conduct periodic reviews of deployed AI features against this policy.

8. Accountability

  • Responsibility for AI governance sits with North's leadership.
  • Concerns about AI behaviour can be reported via the website contact form and are reviewed as part of our AI improvement process.

9. Continuous Improvement

AI regulation and best practice are evolving quickly. We review this policy regularly and update our practices in line with UK, EU and US regulatory developments.