Trust
The AI model providers Kuro AI relies on and how we protect your data.
Kuro AI uses third-party AI models to deliver product features. Every one of them runs inside Amazon Bedrock, configured so that your inputs and outputs are never used to train foundation models and the model vendors never receive your data.
Last updated: September 2026
How we handle AI provider data use.
We do not send your data to model vendors directly. Instead we call every model through Amazon Bedrock, a managed hosting layer where the model runs inside AWS infrastructure we control the region of. That single architectural decision is what lets us make the commitments below.
No training on your data
Every model we use is hosted for us by Amazon Web Services on Amazon Bedrock, whose terms prohibit the use of customer content for model training.
Vendors never receive your data
Because inference runs inside Bedrock, the companies that built these models do not receive your prompts, your documents or your business data.
UK first, EEA only by exception
Chat, analysis and document search run in the United Kingdom. Two features process transiently elsewhere in the EEA because the model we need is not offered in the UK — both are named below.
AI model providers.
The following AI model providers may process customer data when delivering Kuro AI’s product features. Each provider’s terms or acceptable use policy is linked where available.
NVIDIA
Large language model processing for chat, analysis and agent reasoning. Primary model provider.
NVIDIA agreements
Amazon Web Services (Bedrock)
Managed AI model hosting for every model request Kuro makes. Also provides the models behind document search and the voice assistant.
AWS service terms
Cohere
Model inference for search re-ranking. Applies when a search returns enough results to be worth ordering by relevance.
Cohere terms of use
Anthropic
Model inference for selected AI features and provider routing. Used when specific model paths or routing rules select Anthropic.
Anthropic commercial terms
Meta
Model inference for selected AI features and provider routing. Used when configured model routes select Meta.
Meta acceptable use policy
Mistral AI
Model inference for selected AI features and provider routing. Used when configured model routes select Mistral AI.
Mistral AI terms
OpenAI
Model inference for selected AI features and provider routing. Used when configured model routes select OpenAI’s open-weight models.
OpenAI usage policiesRouting and conditional providers.
Kuro routes most work to our primary provider. The others above are conditional: they process your data only when a configured model route selects them, for particular tasks, for resilience, or while we evaluate a better option.
Wherever routing applies, we apply provider-selection controls so customer data only reaches providers whose data-use commitments match the statement at the top of this page: hosted for us by Amazon Web Services, no training on your content, and region-pinned so your data cannot be processed outside the UK or EEA.
Adding a provider does not change where your data goes or who may train on it. If a change to our model providers would affect either, we publish it here and in our privacy policy before it takes effect.
Where models run.
Your data at rest — database, files, cache and backups — stays in the United Kingdom. Chat, analysis and document search also run there. Two features process transiently elsewhere in the EEA because the model we need is not offered in the UK:
Voice assistant — Sweden
While a voice session is open, your audio and the conversation text are processed by Amazon Web Services in Sweden over a private network path, never the public internet. Nothing is stored there and the model does not retain it.
Search re-ranking — Germany
Your question and short excerpts of some documents that matched it are sent to Germany to order results by relevance. Nothing is stored there and the model does not retain it.
Transfers from the UK to the EEA are permitted under the UK GDPR without further safeguards, because the UK’s adequacy regulations cover the EEA. Full detail is in section 8 of our privacy policy.
Scope.
This page covers AI model providers only. Our other processors — hosting, email delivery, payments and AI observability — are listed in section 9 of our privacy policy.
Tools you connect yourself — your accounting software, CRM, point of sale, email, calendar or cloud storage — are governed by your own relationship with those services and their terms. What Kuro reads from each, and whether it can write back, is set out on our features page.
Questions about AI providers or data use?
If you need more detail on any model listed here, or on how we handle your data before signing up, we are happy to talk it through.