AI Provider Statements

title: AI Provider Statements — Training and Data Retention version: 1.0 owner: Pocket PMO Ltd last_reviewed: 13 Jul 26 status: Approved audience: Public (Trust Centre)

AI provider statements — training and data retention

Pocket PMO uses third-party large-language-model (LLM) providers through the Lovable AI Gateway. This document quotes each provider's public commitments about how customer prompts and completions are (or are not) used for model training, and how long they are retained.

We update this page whenever a provider updates its terms. If any statement below becomes out of date, please contact us at admin@pocketpmo.co.uk.

OpenAI (GPT-5 family)

No training on API traffic by default.

"OpenAI does not train our models on inputs and outputs through our API, unless you explicitly opt in to share data with us."

Source: OpenAI Enterprise Privacy — "API, ChatGPT Enterprise, and ChatGPT Team".

Retention

"By default, we retain API inputs and outputs for up to 30 days for the purposes of providing the services and to identify abuse."

Source: OpenAI Enterprise Privacy — "Data retention".

Sub-processing

OpenAI publishes its own sub-processor list (Microsoft Azure infrastructure, etc.) at https://openai.com/policies/subprocessor-list/.

Pocket PMO does not enrol in any OpenAI training programme. All API traffic flows through the Lovable AI Gateway with default (non-training) settings.

Google (Gemini 2.5 / 3.x — via Google Cloud / AI Studio API)

No training on paid-tier API traffic.

"When you use the paid services, Google doesn't use your prompts (including associated system instructions, cached content, and files such as images, videos, or documents) or responses to improve our products, and does not log them with your identity."

Source: Gemini API Terms of Service — Paid Services and Google Cloud Vertex AI — Generative AI Data Governance.

Retention

"Google logs prompts and responses for a limited period of time, solely to detect and prevent abuse or misuse."

The exact retention window is currently up to 55 days for abuse-monitoring logs in the paid Gemini API tier (per Google's linked policy above).

Pocket PMO uses only paid-tier Gemini traffic through the Lovable AI Gateway. No free-tier Gemini endpoints are used from the product.

Anthropic (Claude — where used for internal Lovable tooling only)

Pocket PMO does not currently invoke Anthropic Claude from the customer-facing product. If that changes we will add Anthropic's equivalent statement here. For reference, Anthropic's public commitment is:

"By default, we will not use your inputs or outputs from our commercial products to train our models."

Source: Anthropic Privacy Policy.

Lovable AI Gateway

The Lovable AI Gateway is the routing layer between Pocket PMO and the providers above. Lovable's own statement is that it does not train models on customer data and does not retain prompts or completions beyond what is needed to serve the request and provide observability. See the Lovable AI Gateway documentation for the current statement.

What this means for Pocket PMO customers

• We do not train any AI model on your data, and neither do our providers by default.
• We do not opt in to any training programme on your behalf.
• Retention with providers is short (up to 30 days at OpenAI, up to 55 days at Google) and used only for abuse prevention.
• You can disable AI features entirely for your organisation from Settings → AI, in which case no prompts leave the platform at all.

If you require a signed statement to this effect for procurement, please contact admin@pocketpmo.co.uk and we will countersign our AI Transparency Statement for your organisation.

Version history