AI use policy.

Who provides this service#

SpotRivals is operated by a besloten vennootschap (BV) registered in the Netherlands, KvK 77807049. Full entity identification is on the Legal information page. The provider of the service is the operator under Article 3 of the EU AI Act; the deployer of the AI in the service is the same entity.

What AI we use#

SpotRivals uses one AI model, called server-side from one backend job:

Model

Claude, by Anthropic (the most recent model family at the time of this writing). Called via the Anthropic Messages API.

Where it runs

Anthropic’s infrastructure, primarily in the United States. Transfers covered by Standard Contractual Clauses and Anthropic’s EU-US Data Privacy Framework certification.

What it does

Reads the textual diff between two snapshots of a page you asked us to watch and writes a short strategic-impact reading plus a recommended action.

How often

Once per detected change that survives our noise filter. Most pages produce zero AI calls per week.

There are no other AI features in the product. The marketing site is hand-written. The dashboard is hand-coded. The classification step that decides whether a change is Critical, Important, or Worth noting uses the same Claude call as the analysis itself.

What the AI is allowed to see#

The AI receives a tightly scoped payload for each change:

• The textual diff between the previous and current snapshot of one public competitor page.
• The URL the snapshot came from.
• A short label classifying the page (pricing, changelog, careers, homepage, comparison, blog, trust).
• The date the change was detected.

The AI does not receive your account email, your name, your company, your team members, your billing details, your watchlist composition, your notes, or the contents of any earlier brief.

What the AI does (and does not) produce#

For each change, the AI produces a short paragraph in plain English: what changed, why we think it matters strategically, and what you might consider doing about it. The output appears in two places: as the body of the change card in your dashboard and as the corresponding line in the Monday brief.

The AI does not generate financial projections, legal positions, regulatory advice, or strategic decisions you should treat as authoritative. It is a decision-support layer over change detection. You decide what to do.

Limitations of AI output#

Large language models can be wrong in three ways that are worth knowing about:

1. Misreading the diff. The model can occasionally describe a change that is not really there, or miss the most important change in a noisy diff. The change-detection layer is deterministic; the analysis layer is not.
2. Over-confident framing. The output can sound certain about strategic implications that are, in reality, one of several possible reads. Treat the analysis as a starting point for your judgement, not a substitute for it.
3. Stale context. The model does not know your business, your roadmap, or your sales pipeline. It only knows the diff in front of it. Your context is what makes its output actionable.

The Terms of Service make this binding: AI analyses are best-effort and not financial, legal, or strategic advice. Do not act on them without your own judgement.

Training and reuse#

Anthropic’s API contract for SpotRivals includes Zero Data Retention: the prompts we send and the completions we receive are not stored on Anthropic’s side beyond what is necessary to return the response, and they are not used to train Anthropic’s models. We do not train any model of our own on your inputs or on the AI outputs.

We may use aggregated, de-identified statistics about how often the analysis layer runs, how long it takes, and how often users mark its output as useful, to improve the prompt and the surrounding product. We do not retain or share the contents of any individual analysis for that purpose.

Your responsibility for downstream use#

The output of the AI analysis is yours to use inside your business. You may quote it in internal documents, board updates, sales-team briefings, and strategy meetings. The Acceptable-use policy adds two limits:

• Do not republish AI summaries verbatim as your own writing on a public website, in a blog post, in a press release, or as social-media content.
• Do not represent the AI output as advice from a regulated profession (legal, financial, medical) you do not yourself hold.

You write your own analysis informed by what SpotRivals surfaced; that is the entire point of the product.

EU AI Act#

SpotRivals is an information-society service that uses a general-purpose AI model (Claude) to generate text. Under the EU AI Act this is a limited-risk use of AI: the AI does not make autonomous decisions, it does not interact with physical systems, and it is not used in any of the high-risk domains listed in Annex III of the Act.

This page satisfies the Article 50 transparency obligation: you are told that AI is used, where, and on what data. The output of the AI is clearly labelled in the dashboard (“AI analysis”) and in the Monday brief.

If the EU AI Act regulator (the Dutch Authority for Digital Infrastructure, AT Digi, in cooperation with the Autoriteit Persoonsgegevens) issues further guidance applicable to a service of this kind, we will update this page and notify account owners 14 days before any material change.

Changes to this policy#

Adding a new AI vendor, expanding what data the AI sees, or changing the category of decisions the AI is involved in are all material changes. We will email account owners at least 14 days before any of them go live and bump the version number at the top.

Contact for any AI-related question: privacy@spotrivals.com.