← Back to insights
AI & Law
When a vendor retrains on your data: the risk hiding in default terms
Calling a third-party AI, whether your inputs are used to retrain is often in default terms you didn't read closely.
Many AI services' default terms let the vendor use your input to improve models. If your input contains user personal information or trade secrets, this may breach your promises to users or leak secrets.
The problem is these terms are often on by default, buried in long agreements. Integrating without reading closely makes an unconsented decision for your users and yourself.
The pragmatic approach is to review vendor terms, turn off retraining (or choose an enterprise tier), and reflect the true data flow in your privacy policy and DPA.