Terms and conditions

Starter template — legal review required before public launch. This document describes BrandBanta's service as it is actually built today: the quotas the code enforces, the bring-your-own-key model, and the deletion behaviour that is actually implemented. It is not legal advice and has not been reviewed by counsel. Every commercial term (entity, jurisdiction, governing law, pricing, refunds, liability caps, indemnities) is marked _to be filled at publication_ and must be settled with a qualified attorney in the operating jurisdiction before this page is published or approved: true is set. See the "Open items for counsel" section at the end.

Effective date: to be filled at publication Last updated: 15 August 2026

1. Who these terms are between

These Terms of Service ("Terms") are an agreement between:

BrandBanta — operated by Aleksandar Perisic, a sole natural person (Einzelperson; no legal entity and no commercial-register entry), Waldspielplatz 3, 82319 Starnberg, Germany ("BrandBanta", "we", "us"); see the Imprint, and
you — the individual accepting these Terms, and, where you accept on behalf of a company or other organization, that organization ("you", "Customer").

If you accept on behalf of an organization, you represent that you are authorized to bind it. By creating an account, accessing a workspace, or using the service, you agree to these Terms.

Related documents, each incorporated by reference:

Privacy Policy — what personal data we process and why
AI Policy — how AI is used, and the limits of AI-generated output
Data Processing Agreement — where you are a controller and we process personal data on your behalf

2. What the service does

BrandBanta is a measurement tool. It sends prompts you configure to third-party AI assistants (currently ChatGPT/OpenAI, Claude/Anthropic, Gemini/Google, and Perplexity), stores their responses verbatim, and computes visibility metrics from them — mention rate, share of voice, sentiment, position, and cited sources.

We are a deployer of third-party AI models, not an AI provider. We do not train models on your data. The models are operated by the providers named above, under their own terms.

The service is provided as a hosted web application. Access is authenticated by session; there is no customer-issuable API credential — we do not currently offer programmatic API keys, and the interactive API reference is not exposed in production.

3. Accounts and workspaces

You must provide accurate account information and keep your credentials secure. You are responsible for activity under your account.
Work is organized into workspaces (organizations). A workspace has owners, administrators, and members.
Workspace administrators can see workspace-level metadata, including scan counts and inference cost attributed to each member. They cannot see, via the cost dashboard, the text of prompts or AI responses attributed to an individual member. This is described in detail in the Privacy Policy.
Workspace owners and administrators control provider credentials, members, and settings for that workspace. If you join a workspace created by your employer or client, that organization — not you — controls the workspace's data.
You must be at least 16 years old. The service is intended for business use and is not directed at children.

4. Plans, quotas, and fair use

4.1 Launch is free-only

Paid billing is not enabled at launch. Unless and until we enable it and you expressly subscribe, the service is provided free of charge and the paid terms in §4.4 do not apply to you.

4.2 Limits that apply to free use

Two independent limits apply, and the stricter one governs:

| Limit | Value | Behaviour | | ---------------------------------------- | ----------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------- | | Workspace scan quota | 75 scans per month | Resets monthly. | | Scans run on our shared provider key | 5 scans in total, for the lifetime of the workspace, and at most 2 per hour | Does not reset. Once exhausted, scans require you to connect your own provider key (§5). |

In other words: the free plan is bring-your-own-key. A workspace receives five scans in total on our shared key — enough to set up — after which running scans requires you to connect your own provider key (§5); with your own key connected, the free plan allows up to 75 scans per month. The shared-key limit exists because those scans are paid for out of our own provider account.

Scans run on our shared key are also processed in a reduced mode: output length is capped, the second-pass AI extractor (sentiment, position, and context classification) is skipped in favour of deterministic mention detection only, and model selection is fixed to defaults.

We may also apply per-workspace spend caps and request rate limits to protect the service and to prevent runaway inference cost. When a limit is reached, the request is refused with a clear error rather than silently degraded.

4.3 Changing limits

We may change quotas, rate limits, and free-tier allowances. For changes that materially reduce what a workspace already relies on, we will give at least 30 days' notice in-product and by email to registered users.

4.4 Paid plans — to be settled with counsel before billing is enabled

The following must be filled in and reviewed before paid billing is turned on. Do not publish prices here until the pricing page, the documentation, and the payment configuration state the same numbers.

Plan names, quotas, and prices — to be filled at publication
Billing period, currency, and taxes — to be filled at publication
Trial terms — to be filled at publication
Renewal, cancellation, and whether partial periods are refundable — to be filled at publication
Upgrade/downgrade and proration behaviour — to be filled at publication
What happens to data that exceeds a lower plan's allowance after a downgrade — to be filled at publication

Payment processing, if enabled, is handled by Stripe. We do not receive or store full card details.

5. Your provider keys (BYOK)

You may connect your own credentials for supported providers (OpenRouter, Anthropic, OpenAI, Google, Perplexity, and Tavily) in workspace settings.

You pay the provider directly. When a workspace's own key is used, the inference is billed to your account with that provider, under your agreement with them. We do not mark up or resell inference on those runs.
That provider becomes your processor for the workloads you send through your key. Their terms, availability, and data-handling practices govern that traffic; we are not responsible for them.
Keys are encrypted at rest and decrypted only at the moment of use. We display only the last few characters. Plaintext keys are never logged.
We do not silently substitute our key for yours. If your key cannot be used, the affected scan fails with an error telling you to re-enter the key; it does not quietly fall back to our shared key and our expense. Certain execution modes (direct-to-provider and batch) run only on your own key and are skipped entirely if one is not available.
You are responsible for keeping your provider account funded and in good standing, and for any charges incurred through your key — including charges resulting from scans you or your workspace members schedule.
Remove a key at any time in workspace settings.

6. Acceptable use

You must not:

Use the service in violation of applicable law, or in violation of the terms of any AI provider whose models the service calls
Attempt to circumvent quotas, rate limits, spend caps, or the free-tier allowance — including by creating multiple workspaces or accounts to obtain additional free scans on our key
Probe, scan, or test the vulnerability of the service, or interfere with its operation, other than through a good-faith security disclosure (§15)
Access another workspace's data, or attempt to
Submit prompts designed to elicit unlawful content, or to harass, defame, or impersonate a person
Resell, sublicense, or provide the service to third parties as a service of your own, except as expressly agreed in writing
Use the service to build a competing product, or to scrape it in bulk

We may suspend a workspace that we reasonably believe is engaged in the above, or that is generating cost or load that threatens the service for others. Where practical we will contact you first.

7. Your content

You keep ownership of what you put in: your brand names, topics, prompts, aliases, and workspace configuration.
You grant us a limited licence to host, process, and transmit that content for the sole purpose of operating the service for you — including sending your prompts to the AI providers you have enabled.
AI responses are stored verbatim so you can audit what a model actually said. Those stored responses, and the metrics derived from them, are part of your workspace's data.
We do not train models on your data, and we do not sell it. See the AI Policy for what the providers do with data sent to them.
You are responsible for ensuring you have the right to submit what you submit, and for not putting personal data into prompts unnecessarily.

8. AI output — accuracy and reliance

This is important, and it is the nature of the product.

BrandBanta reports what AI assistants said, not what is true. AI models hallucinate, contradict themselves, reflect the biases of their training data, and change behaviour when providers update them. A response may state something about your brand — or a competitor's — that is simply false.

Metrics are computed from a sample of model responses at particular times, not from a census. They are indicative, not exhaustive.
Sentiment, position, and context classifications are best-effort machine enrichment of the stored text. The stored text is the source of truth.
Narrative summaries and recommendations are advisory. You remain the decision-maker.

Do not rely on the service's output as fact without independent verification, and do not use it as the sole basis for regulatory disclosures, financial decisions, or public claims about a competitor. Methodology is documented in the product documentation and on the methodology page.

9. Availability

The service is provided as-is and as-available. We do not currently offer an uptime commitment, service level agreement, or guaranteed support response time, and nothing in these Terms or in our documentation should be read as one.

We may modify, suspend, or discontinue features. For changes that remove functionality a workspace relies on, we will give reasonable notice.

Scheduled scans, alerts, digests, and other background processing depend on third-party AI providers and infrastructure. Provider outages, rate limits, and model deprecations can delay or prevent scans, and are outside our control.

10. Suspension, termination, and deletion

10.1 You can leave at any time

You may delete your account, or delete a workspace you own, from the settings pages.

Deletion is immediate and irreversible. There is no grace period, no soft-delete, and no restore path. Deleting a workspace removes its scans, responses, metrics, and stored provider credentials. Deleting your account removes your user record. Export anything you want to keep before you delete.

Backups exist for disaster recovery, not for restoring individual deletions, and are themselves cycled out.

10.2 We can terminate

We may suspend or terminate access for material breach of these Terms, for non-payment (once billing is enabled), or where required by law. Where practical we will give notice and an opportunity to fix the problem.

10.3 What survives

Sections on your content ownership, AI-output disclaimers, limitation of liability, indemnity, and governing law survive termination. We may retain records we are legally required to keep; see the Privacy Policy.

11. Disclaimers and liability — to be settled with counsel

The operative wording of this section — warranty disclaimers, exclusion of indirect and consequential loss, the liability cap and its measure, and any carve-outs (death or personal injury, fraud, gross negligence, and other liability that cannot lawfully be excluded in the governing jurisdiction) — must be drafted or approved by counsel. Consumer-protection law in some jurisdictions overrides broad exclusions.

The intended commercial position, for counsel to render into enforceable terms:

The service is provided as-is, without warranties of merchantability, fitness for a particular purpose, or non-infringement.
We do not warrant that AI output is accurate, complete, or fit for any decision you make with it.
We are not liable for the acts, omissions, pricing, availability, or data handling of third-party AI providers, including where you use your own key with them.
Liability cap — to be filled at publication. Note that at launch the service is free of charge, which counsel should account for in setting the measure.

Indemnity — to be filled at publication.

12. Changes to these Terms

We may update these Terms. The "Last updated" date reflects the most recent revision. Material changes will be announced in-product and by email to registered users at least 30 days before they take effect. Continuing to use the service after that date means you accept the revised Terms. If you do not accept them, stop using the service and delete your account.

13. Governing law and disputes — to be settled with counsel

Governing law — the laws of the Federal Republic of Germany, excluding its conflict-of-law rules and the UN Convention on Contracts for the International Sale of Goods (CISG). Mandatory consumer-protection law of a consumer's country of residence is unaffected.
Exclusive jurisdiction / venue — to be filled at publication
Whether disputes go to arbitration, and any class-action waiver — to be filled at publication (enforceability varies sharply by jurisdiction, and consumer rules may make some choices unavailable)
Mandatory consumer-law rights that cannot be displaced — to be confirmed for each market where the service is offered

14. General

Entire agreement — these Terms, with the Privacy Policy, AI Policy, and (where signed) the DPA, are the whole agreement between us on this subject.
Severability — if a provision is unenforceable, the rest stands.
No waiver — not enforcing a term once does not waive it.
Assignment — you may not assign these Terms without our consent; we may assign them to a successor in connection with a merger, acquisition, or sale of assets.
Notices — we will contact you at the email address on your account. Contact us at the addresses below.

15. Contact

General and commercial — hello@brandbanta.com
Privacy — privacy@brandbanta.com
Security disclosures — security@brandbanta.com

Postal address: Aleksandar Perisic, Waldspielplatz 3, 82319 Starnberg, Germany

Open items for counsel

This draft was written from the implemented behaviour of the software. The following were deliberately not asserted, because the code does not currently support them or because published surfaces disagree with each other. Each needs an owner decision before publication:

Price. Three surfaces state different numbers for the paid tiers. No price is stated here until they agree.
Refunds, proration, and downgrade behaviour. Described in the product documentation, not implemented in code.
Data retention periods. The Privacy Policy commits to retention windows for scan results and audit logs that no code currently enforces. Either implement them or amend the commitment — do not restate them here in the meantime.
Per-tier history retention. Advertised on the pricing surfaces; not implemented in either direction.
"Paid plans require your own key." Stated in the documentation; not enforced in code.
Seat limits. Seats are counted for billing but never capped.
Legal entity, address, jurisdiction, governing law, effective date.
Liability cap, indemnity, and dispute resolution.
Whether a separate DPA must be executed with each business customer, or incorporated by reference here.