AI Safety

Tidal Grass builds software and AI features that we would be comfortable letting our own families use. This page explains how we develop, deploy, and oversee AI across our products so the people who rely on us can trust what they are using.

Our principles

• Helpful by design. AI features exist to assist people, not to replace their judgment. We build them to support clear, well-understood tasks.
• Honest and transparent. We tell you when a feature uses AI and we do not disguise AI-generated content as human work.
• Safe and accountable. We test for harm before release and we keep humans responsible for outcomes.
• Private by default. Your data is not used to train models without your consent, as described in our Privacy Policy.

How we use AI

We use AI in a focused set of places where it adds real value:

• Assisting with writing, summarizing, and organizing information inside our products.
• Helping our support and engineering teams understand issues and respond faster.
• Improving accessibility, search, and navigation across our sites and apps.

We do not use AI to make decisions that significantly affect a person’s legal, financial, employment, or medical situation without a human reviewing the result.

Model providers

Where an AI feature requires a model, we work with established providers and select models based on safety, reliability, and suitability for the task. Our provider relationships are governed by agreements that address confidentiality, acceptable use, and data handling. When a provider processes data on our behalf, it is only to deliver the feature you requested.

Safeguards

Before an AI feature ships, and continuously afterward, we apply safeguards including:

• Review and testing. Features are reviewed for misuse, bias, and failure modes before launch, with prompts and outputs tested against our safety criteria.
• Content guardrails. We configure input and output filters appropriate to the feature and its audience.
• Human oversight. People review flagged outputs and can intervene, override, or escalate.
• Monitoring. We monitor usage and feedback to catch problems early and improve over time.
• Incident response. When something goes wrong, we investigate, fix it, and document what we learned.

Data handling

• We minimize the data sent to AI features and only include what is needed for the task.
• We do not use your content to train models without your consent.
• Provider data handling is bound by confidentiality and purpose limitations, as described in our Privacy Policy.
• Inputs and outputs are retained only as long as needed to operate, support, and improve the feature, or as required by law.

Contact

For questions about how Tidal Grass uses AI, or to report an AI safety concern, email us at hello@tidalgrass.com.