Responsible AI & Acceptable Use

Radarist helps people make better decisions from evidence. It is a decision-support tool, not a decision-maker. These principles describe how to use it well — and how not to.

Our design principles

Evidence over assertion. Relations and claims are meant to carry sources. Treat unsourced output as a lead, not a fact.
Human-in-the-loop by default. Agents propose; people decide. Autonomy is earned and measured, never assumed. See the thesis.
Calibrated, not overconfident. Confidence scores are estimates. The system is designed to abstain and escalate when it is not sure enough.
Local-first and private. Your data stays with you unless you choose otherwise.

Use it for

Technology scouting, competitive and market intelligence, and innovation decisions.
Organizing evidence into a knowledge graph you own and can interrogate.
Drafting analysis and reports that a human reviews before they are acted upon.

Do not use it for

Fully automated, unreviewed decisions with legal, financial, safety, or human-rights impact.
Surveillance of individuals, harassment, or any unlawful purpose.
Generating or laundering disinformation, or misrepresenting AI output as verified fact.
Ingesting data you do not have the right to use, or violating third-party terms.

Your responsibilities

Because Radarist runs under your control, you are responsible for the data you ingest, the third-party AI providers you connect, and the decisions you make from its output. Verify important claims against their sources before relying on them.

Reporting concerns

To report a concern about misuse or AI safety, contact conduct@radarist.ai. For security issues, see our Security Policy.