AI and Automated Processing Notice
In short
- Automated processing helps read pitch decks, match records and summarise evidence.
- It never decides anything on its own, and it never invents a source.
- Every finding is tied to a named source and a date you can inspect.
- We do not use your data to train general-purpose AI models.
1. Where automated processing is used
| Where | What it does |
|---|---|
| Pitch-deck and business-plan reading | Extracts the factual claims a document makes — roles, dates, education, traction, intellectual property — so they can be checked |
| Record matching | Decides whether a public record belongs to the founder, using name variants together with date of birth and country of residence |
| Prior-art and similarity screening | Surfaces patents and publications close to the technology described |
| Summarising | Turns raw source output into the short line shown on each ledger row, and drafts the founder-market-fit analysis |
| Founder-market fit | Produces a written assessment of the founder's background relative to the market described |
The founder acknowledges these uses in the versioned consent set submitted with the intake form. Continuing from the introductory privacy notice to the form does not itself record consent. Automated processing starts only after the complete consent set and founder information are successfully submitted.
2. What it does not do
- It does not decide whether to invest, and produces no score, rating or recommendation.
- It does not produce a finding without an underlying source.
- It does not conclude anything from silence: a source that is down or out of coverage is reported as Unavailable or Pending, never as a clear result.
- It does not attribute a record to a person on a name match alone.
3. Human oversight and inspectability
Every automated output is anchored to a source label, a date and, where the record is public, a link. The subscriber sees the evidence, not just the conclusion, and can inspect the underlying record. Founders see the report about them and can comment on or dispute any line.
We review matching and classification logic regularly, and test it against known cases, with the deliberate bias of leaving a line unmatched rather than attaching the wrong person.
4. No solely automated decisions with legal effect
We do not make decisions producing legal or similarly significant effects on an individual by automated means alone. Investment and business decisions are taken by the subscriber, exercising their own judgement, based on evidence they can inspect.
4a. EU AI Act and Quebec Law 25
Under the EU Artificial Intelligence Act, the automated components described above are used to assist a human diligence process; they do not evaluate creditworthiness, do not assess a natural person for employment, education, housing, insurance or access to essential services, and produce no score or ranking. Where the Act imposes transparency obligations on AI-generated text, the ledger marks AI-drafted summaries and analyses as such.
Under Quebec's Law 25 (Act respecting the protection of personal information in the private sector, s. 12.1), we inform founders that automated processing is used to match records and draft summaries; any founder may ask which personal information was used, the principal factors and parameters behind a line, and may have the line reviewed by a person and corrected. The right to an explanation and to contest applies to every founder regardless of where they live.
5. Limits of automated output
Language models can mis-summarise, miss nuance or over-generalise. Extraction from a document can misread ambiguous layouts. Similarity screening indicates overlap for review; it is not a patentability or freedom-to-operate opinion. Read the source, not only the summary. See No Warranty, Limitations and Proper Use.
6. Data used for AI processing
Document text and verification results are sent to AI processing providers under contract, limited to producing your output. We do not permit them to use it to train general-purpose models, and we do not do so ourselves. Categories of provider are listed on the Sub-processors page.
7. Your rights
You can ask what automated processing was applied to a report about you, ask for human review of a line, and dispute it under the Accuracy, Correction and Dispute Policy.