Legal
AI Disclosure
What Caspr is, how it produces an analysis, and how its outputs are marked.
Effective 2 August 2026 · Last updated 22 August 2026
This page states what Caspr is, how it produces the analyses it delivers, and how those outputs are marked. It is published in line with Article 50 of the EU AI Act and applies to every Caspr account, on every plan, including the free trial.
Caspr Is an AI System
Every analysis Caspr delivers is generated by an AI system. No part of a Caspr report is written by a human analyst on our staff, and none of it should be read as human authorship.
Caspr is not a general-purpose assistant. It is purpose-built for business analysis: it takes a single question, proposes a structure, asks what it needs to know, and returns a structured, cited document.
How an Analysis Is Produced
Two systems do the work. The Thinking Brain is the research methodology — it decomposes a question, decides what evidence would answer it, weighs alternatives against each other, and structures the conclusions. It reasons through the material rather than retrieving passages from it.
The Learning Brain supplies current data. Caspr reads live feeds rather than relying only on what a model absorbed during training, so an analysis reflects the state of the world on the day it is run.
Between them, a report passes through model reasoning, source retrieval, verification against the retrieved source, and document assembly. A human is not in that loop.
Where the Sources Come From
Caspr draws on a curated corpus of more than 25 million sources — company filings, government and multilateral databases, academic papers, news feeds and industry reports. It is a curated corpus, not an open crawl of the web.
Every insight in a report is shown with the source it came from. Where a claim cannot be attributed to a credible source, it does not appear.
How Outputs Are Marked
Every document Caspr generates carries its provenance in two forms.
Visible — the cover and the document footer of every PDF and PPTX state: Generated by Caspr · Analytical AI · [date].
Machine-readable — the file's own metadata records that the content was AI-generated, when it was generated, at what depth, and how many sources it drew on. It is embedded in the file, so it travels with the document, and it is re-applied whenever a report is regenerated after editing.
These marks are not conditional on your plan. Free-trial output carries them identically.
What Caspr Does Not Mark
Documents you upload to a Data Room are yours. Caspr did not generate them and does not mark them. Only the deliverable Caspr produces is marked.
Limits
Caspr is built to be defensible: every source credible, every claim triangulated, every conclusion attributed. It is not built to be infallible. An analysis is evidence for a decision, not the decision itself.
Caspr does not provide investment, legal, tax or medical advice, and its output is not a substitute for professional advice in any regulated field.
Your Data and Model Training
Your prompts, your uploaded documents and your reports are never used to train any model, ours or a third party's. Caspr is ISO 27001:2022 certified and GDPR compliant; the full position is set out in our security posture and privacy policy.
Contact
Questions about this disclosure, or about how a particular report was produced, go to support@caspr.ai.