Engines and confidentiality
Your organisation's engine policy: the engines allowed, the notice displayed before text is sent to a foreign engine, and the use of your own API keys.
How do we choose which engines our users can access?
Each organisation has its own list of allowed engines (‘Allowed AI providers’). This is the organisation's engine policy: a department that handles sensitive matters can, for example, restrict itself to LexMachina.
- In Highly sensitive content mode, only Swiss engines can be selected, whatever the list.
- Who does this: the Neur.on team, during set-up or on request.
How do we display a notice before text is sent to a foreign engine?
The ‘Show confidentiality warning when non-confidential engines are selected’ setting displays the Attorney-Client privilege notice as soon as a user chooses an engine hosted abroad. It is enabled organisation by organisation and is not passed on to child organisations.
- The notice reminds users of your organisation's policy when they choose the engine, in the tool itself.
- Who does this: the Neur.on team, during set-up or on request.
Can we use our own DeepL, Claude or GPT contracts?
Yes. By default, Corrext uses its own access keys for these engines. An organisation can provide its own API keys for DeepL, Claude and GPT, so that requests go through its own contracts with these providers.
- Processing by these engines still takes place outside Switzerland: the Attorney-Client privilege notice applies in the same way.
- Who does this: the Neur.on team, during set-up or on request.