Parliamentary casework is often deeply sensitive. Offices deal with immigration problems, health issues, financial hardship, safeguarding concerns, and family crises.
That is exactly why we thought carefully about how AI should be used inside ParlyReply.
We want parliamentary teams to be able to use genuinely useful AI features without lowering the standard on constituent data. That is why we secured a Zero Data Retention arrangement with OpenAI for the AI features inside ParlyReply.
In practice, that means casework sent to OpenAI is processed to generate a response, but is not stored by OpenAI, kept in long-term logs, or used to train future models.
If an office wants to review the agreement itself, it is available on request.
We pushed for this because we think the usual trade-off offered by AI tools is the wrong one for parliamentary work. Offices should not have to choose between better software and careful data handling.
Zero Data Retention is not the whole story. Sensitive casework still needs proper access controls, office-level separation, and a clear audit trail. But it does remove one important concern: the model provider is not retaining the underlying casework.
Inside ParlyReply, we still store the information needed to make the product work properly and maintain accountability. That includes things like final drafted outputs, relevant metadata, and audit information showing how work was carried out.
More broadly, this reflects how we think about AI in ParlyReply: the technology should fit the standards already expected in parliamentary offices, rather than asking offices to make exceptions for the sake of convenience.