Understand the concern.
Start a useful conversation.

Who is this for?

Anyone who wants to ask better questions about AI: patients and families, health teams, parents, educators, community groups, and people building or buying tools. You do not need a technical background.

Choose what you need

Explore the concerns map ↗

Browse 1,000 named concerns. Choose a topic, follow a connection, or search for a specific question.

Find questions ↗

Choose your role and situation to find questions you can bring to a visit, meeting, or discussion.

Plan a conversation ↗

Use a guide to help a group listen, discuss a concern, and decide what to do next.

Explore health & care ↗

Look at concerns in clinics, homes, and communities.

Find sources ↗

Go directly to research, reported incidents, frameworks, and guidance from other organizations.

How the explanations are being expanded

1,000 of the 1,000 concerns now have an individual definition, two explanatory paragraphs, an example, a conversation question and linked sources. These expanded guides are working drafts: editorial review and source checks continue.

Sources include research, official guidance, documented cases and policy arguments. Examples are illustrative unless identified as a documented case. A source link is not proof that every possible version of a concern has occurred.

What does the map show?

The register groups concerns into 23 topics. Color identifies the broader group: technology and systems, people and society, or governance and power.

Solid lines show how entries are grouped. Dotted lines show related entries in the register. They do not establish that one concern causes another. A node’s size or position is not a measure of risk.

How should I read the scores?

Each concern has a severity score from 1 to 5 and a likelihood score from 1 to 5. Multiplying these gives the register’s risk score, from 1 to 25. A time label indicates whether the concern is described as current or further ahead.

These are estimates in the register, not measured probabilities or predictions for your situation. Use them to start a discussion, then consider the people, setting, evidence, and safeguards involved.

What guides the project?

• Be clear. Explain the concern in ordinary language.
• Be proportionate. Consider both potential benefit and harm.
• Listen. Include the people affected by a tool or decision.
• Keep learning. Treat questions as a starting point for useful action.

Our cooperative and collaborators

AI Concerns™ is part of the doáb.ai™ cooperative and is developed in partnership with the Institute for AI Translation.

Source organizations retain ownership of their own work. Linking to MIT, WHO, NIST, or another publisher does not imply that they endorse this project.

Names, content, and credits

AI Concerns™ and doáb.ai™ are the brand names used on this site. © 2026 AI Concerns™ for original site content and design. Third-party names and materials belong to their respective owners.

The interactive map uses Constellation, designed by The Thread, under its MIT license. The Figtree typeface is used under the SIL Open Font License. Open-source license terms continue to apply.