Acceptable Use Policy
What you may upload, what you may ask, and what you may do with the answers.
Last updated 14 September 2026
What Sourceteller is
Sourceteller is a business question-answering tool. It reads documents and websites you supply and answers questions from them, citing the passage each claim came from.
It is deliberately narrow. It is not a general-purpose chatbot, not an image or video generator, not a companion or roleplay product, and not a tool for producing content unconnected to your own material. The assistant declines when your documents do not support an answer — that refusal is the product working, not failing.
This policy forms part of our Terms of Service. It applies to you, to everyone you invite into your workspace, and to anyone reaching your assistant through a widget or phone line you publish.
Content you upload
You may only add material you have the right to use.
You must not upload
- Anything you do not own or have permission to use, including copyrighted material belonging to someone else
- Material obtained by breaching a confidentiality agreement, or by scraping a site whose terms forbid it
- Personal data about people who have not been told it is being processed, where your law requires telling them
- Special-category personal data — health records, biometric data, financial account credentials, government identity numbers — unless you have a lawful basis and the safeguards your law requires
- Malware, or files crafted to attack the system that reads them
- Content that is unlawful where you or your customers are
Crawling websites
Point us only at sites you own or are permitted to index. We respect robots.txt by default. Do not use the crawler to copy a competitor's site, harvest contact details, or evade a paywall.
How you may use the assistant
Do not try to break the grounding
The value of the product is that answers come from your documents and are traceable to them. Do not attempt to defeat that — through prompt injection, jailbreaking, hidden instructions embedded in uploaded files, or any attempt to make the assistant answer from outside your material or ignore its instructions.
Do not use it as a general model endpoint
Credits pay for answering questions about your content. Do not use the assistant or the API as a cheap general-purpose language model, to resell model access, to generate bulk unrelated content, or to route another product's traffic through us.
Do not use it against people
- No harassment, stalking, doxxing, or building profiles of individuals without their knowledge
- No impersonating a real person or organisation
- No generating misinformation, or material designed to deceive about who is speaking
- No content sexualising anyone, and nothing involving minors under any framing
- No promotion of violence, self-harm, terrorism, or illegal weapons and drugs
Do not attack the service
- No probing, scanning or penetration testing without written permission
- No attempt to reach another workspace's data, or to bypass workspace isolation
- No overwhelming the service with automated traffic, and no evading rate limits or credit metering
- No reverse-engineering the service or extracting its prompts and configuration to rebuild it
What you may do with the answers
Answers are generated by an AI model reading your documents. They are grounded and cited, and they can still be wrong.
Tell people they are talking to an assistant
If you publish the widget or a phone line to your own customers, you must make clear they are talking to an automated assistant, not a person. Do not configure it to claim to be human. Some jurisdictions require this; it is good practice everywhere.
Do not present answers as professional advice
Answers must not be passed off as legal, medical, financial, tax or other regulated professional advice. If your documents cover a regulated area, a qualified human must review anything that reaches a customer.
Keep a human in the loop for consequential decisions
Do not use answers as the sole basis for a decision that materially affects someone — employment, credit, insurance, housing, medical care, or a legal position. Automated decisions like these are restricted by law in many places.
Check the citation
Every answer names the passage it came from. Before relying on one that matters, open it. That is what the citations are for.
Prohibited business uses
Sourceteller may not be used to build or operate:
- Adult, sexual or companion-chat products of any kind
- Gambling, betting, or trading-signal and investment-advice services
- Essay mills, homework services, or tools sold for academic dishonesty
- Multi-level marketing, pyramid schemes, or any get-rich-quick offer
- Spam operations, cold-outreach farms, or engagement manipulation
- Surveillance of employees, partners or the public without their knowledge
- Anything unlawful under the laws that apply to you or your customers
How this is enforced
We do not read your documents or conversations for interest, and there is no routine human review of workspace content. We investigate when a report reaches us, when automated abuse signals fire, or when we are legally required to.
Where something is wrong we will normally contact you first and give you a chance to fix it. For serious or unlawful breaches we may suspend or close the account immediately. Unused credits and parse pages are forfeited on closure for breach.
Report abuse to support@sourceteller.com. Include enough detail to identify what you saw. Security vulnerabilities go to the same address — see our security page for what to expect.
Changes
We update this policy as the product and the law change. The current version is always here with its date. Account owners are emailed before anything material takes effect.