Trust

AI suggestions are not verified facts.

AI can produce plausible but incorrect answers. A useful recommendation needs context, supporting evidence, and a person able to challenge it.

What is communicated today

This site labels product stages and illustrative workflows, and distinguishes planned capabilities from current public evidence. It does not publish an AI accuracy benchmark or claim autonomous, reliable decision-making.

Do not use an unverified recommendation as the sole basis for a financial, security, legal, or other consequential decision.

Intended safeguards

Our design principles include showing sources and uncertainty, keeping sensitive actions subject to approval, and checking outcomes. These are goals, not confirmation that every safeguard is implemented in every product.

Evaluation should examine unsupported answers, missing evidence, stale inputs, and whether a reviewer can reject a suggestion. Research questions and product status pages provide context, not validation results.