AI notice

You are using an AI service. Every Moat Index reading, explanation and summary on MoatDefender is produced with AI models, under MoatDefender's proprietary, versioned method (what it measures and its limits are described at /methodology). AI-written explanations are marked as such in the app.

1. Which AI is involved

Readings are built with models from AI model providers, including Anthropic and OpenAI. The models see the material you submit or facts extracted from it. The Moat Index number itself is computed by our own code, not written freely by a model. Every new rating also gets an independent quality check, which can lower the confidence label but never changes the number. The providers are listed on Sub-processors.

2. Limits you should know

• The reading depends on the material. Missing evidence lowers the reading; claims without supporting evidence count for less.
• Models make mistakes, and can misread a document. Each reading lists the facts it used so you can check them.
• The reading is an opinion about a company, not investment advice and not a prediction of returns.
• The same evidence gives the same reading; a new reading needs new evidence.

3. Human oversight

• Readings where the independent check disagrees are flagged "review" and carry a lower confidence.
• You can contest a reading at hello@moatdefender.com. We look at the evidence and the method, and if the method was misapplied we say so and re-issue under the next method version. We never change a number by hand.

4. EU AI Act

MoatDefender is an AI system that interacts with people and generates text, so it is subject to the transparency rules in Article 50 of the AI Act (Regulation (EU) 2024/1689): this notice and the labels in the app meet them. We assess that it is not a high-risk system under Annex III: it rates companies, and does not evaluate the creditworthiness of natural persons, employment, education or access to essential services.