How we handle your data.

Where requests are processed

Every request is processed by infrastructure partners based in the European Union. We route only to partners on our sub-processor list, which is part of the data processing agreement we sign with every customer, and we never route to a partner outside the EU, including when we switch to a backup partner.

What we store

We don't store the content of your requests or the model's answers. We store metadata needed for billing and support: time, model, token counts, cost, status and latency. Usage metadata is kept for 24 months for billing and tax purposes.

What our partners store

Most requests are processed with zero data retention: our main infrastructure partners don't store request content, and where a partner lets us choose zero-retention providers, we do. One partner we use only as a backup may keep request content for up to 30 days for abuse monitoring until our zero-retention request is approved; it is marked on the sub-processor list.

Training

We never use your data to train models, and the partners we use don't train on it either: their terms prohibit it, or we have switched training off.

Your agreements with us

Our data processing agreement (DPA) under Article 28 GDPR is part of our terms, and you can download a copy from your console.

Model origin

Some models are made by companies outside the EU, including in China and the US. The model's origin doesn't change where your data is processed: the model is run by our EU-based partners, and the model maker doesn't receive your requests.

Questions

Email hello@powerspring.eu.