How Threada dey use your data with AI
Threada dey use AI take turn your request enter answer wey get evidence and action wey get governance. Dis page dey talk am clear how we dey use your data for dat process — and where di binding terms dey.
Wetin dey come na di short version wey you fit cite of how we dey handle your data with AI: wetin we dey do and wetin we no dey do with Customer Content, where we dey index am, how long we dey keep am, which model provider dey involved, and which third-party attestation we dey publish now. Every point dey link to di canonical page abi contract wey get di full detail.
We no dey train foundation model with your content
Threada no dey train foundation model with Customer Content. Dis na binding commitment for our Terms of Service (§4, Customer Content), no be just talk: you still own your content, and we dey process am under one limited license just to provide and make di Service better for you.
Retrieval, no be training
We dey index your content for retrieval, no be for training. When Threada dey answer, e dey retrieve di passage wey relevant from di source wey you don approve, and e dey ground di response for dem with citation. Di index dey just to find and cite your knowledge for answer time; we no dey use am train shared model again.
Retention wey you dey control
You fit configure retention for each workspace. Work history and analytics, configuration, and audit record, each one get im own retention window, and when contract don end we go delete customer data within 90 days unless law say make e stay longer. Di Data Processing Addendum and Privacy Policy talk di specifics.
Model provider dem
Threada dey use named model provider for inference and embedding — for now na OpenAI and Google (Gemini) — and each one na subprocessor under correct data-protection terms. Di current list, with each provider purpose and location, dey for di subprocessors page, and we dey update am with advance notice if any big change happen.
Certification and attestation
Threada never publish third-party security attestation like SOC 2 Type II abi ISO 27001 for now. Wetin we dey publish and dey keep current na di underlying control evidence — our security architecture, threat model, subprocessor list, and Data Processing Addendum — so di control wey dey behind our claim fit dey inspect today. If your review need attestation status, reach us make we give you di latest documentation.
Question wey people dey always ask
Threada dey train AI model with our data?
Dem dey use our content take improve one shared model wey other customer go enjoy?
How long una dey keep our data?
Which AI provider Threada dey use?
Una get SOC 2 abi ISO 27001?
You need am for paper for one review?
We fit give you di current data-handling and security documentation make you fit do your assessment.