Trust & Transparency
AI Transparency
We do not train AI models on your data.
Published 22 April 2026 · Steleo Publishing Limited · Company No. 11891029 · ICO Registration ZB969283
The short version
BundleCreator never uses your documents, bundles, or case details to train AI models — ours or anyone else’s. Where we do use AI, it is for specific features under strict privacy controls. We do not sell customer data, and we do not share it with any third party for training, advertising, or analytics.
Every AI service we use, and what it does
1. Mind Map — Anthropic Claude API
The Mind Map feature generates a visual index of your bundle by asking Claude to summarise the structure of extracted document text. The Claude API is configured with zero data retention: Anthropic does not store or retain prompts or responses, and your text cannot be used for training, evaluation, or human review.
Before any text reaches Claude, BundleCreator removes personally identifiable information directly in your browser. The client-side redaction covers:
- UK postcodes (for example SW1A 1AA, M1 1AA)
- Phone numbers — mobile, landline, and international
- Email addresses
- National Insurance numbers
- Dates of birth in multiple formats
- Bank sort codes and account numbers
- Court case reference numbers
- NHS numbers and passport numbers
- UK driving licence numbers
- Common personal name patterns found in witness statements and court orders
Only the anonymised, structure-level text leaves your browser. Raw PDF files are never sent to Claude.
2. Read Aloud — Google Cloud Text-to-Speech
The Read Aloud feature converts the text you select into audio using Google Cloud Text-to-Speech. This is a non-generative service — it synthesises speech from text and does not train on content. Audio is returned to your browser for playback and is not stored on our servers.
3. Document OCR — Tesseract (in your browser)
When a scanned or image-based PDF needs optical character recognition, BundleCreator runs Tesseract entirely inside your browser. The document bytes are never uploaded to a cloud OCR service. Your scanned pages do not leave your device for OCR processing.
4. PII Redaction — in-browser regex (no AI)
The PII redaction that runs before the Mind Map AI call is a set of pattern-matching rules executed in your browser. It does not use a machine-learning model and it does not make any network request.
AI services we do not use
We do not currently use Google Gemini, OpenAI (ChatGPT or GPT APIs), Mistral, Cohere, Perplexity, or any other generative AI provider. If this ever changes we will update this page and our privacy policy before any customer data is processed by a new provider.
What customers keep asking
Could my document text be used to improve Claude?
No. Under the zero-data-retention configuration we use, Anthropic does not retain or log prompt or response content at all. Material that is not retained cannot be used for training, fine-tuning, or human review.
What if a sub-processor changes its data-use policy?
Our sub-processors are bound by the data-processing agreements in place at the time of each request. If Anthropic, Google, or another provider changed its default terms, we would migrate away or renegotiate before accepting any weaker protection. Where zero data retention is available, it is a hard requirement — not an optional setting.
Do you use AI to make decisions about me?
No. BundleCreator does not use AI for pricing, eligibility, risk scoring, account suspension, or any other decision that affects you. The AI features on the platform are user-triggered and assist with document organisation only.
Keeping us honest
This page is maintained alongside the code that implements each feature. If you believe anything on this page is out of date or inconsistent with how the product actually behaves, please contact us at enterprise@bundlecreator.co. Security or privacy concerns can also be reported under our Vulnerability Disclosure Policy.