Privacy protects information and confidentiality associated with people who use or receive a service. In LLM-assisted work, it requires protecting customer and user data, confidential source code, meeting notes, and organizational intellectual property from uncontrolled disclosure. section 2.2.3
Apply privacy when deciding what to send to an LLM:
- Check organizational policies, nondisclosure obligations, and laws governing user data before sharing information. section 2.2.3 section 6.2.2
- Choose prompts and LLM deployments according to what data may be shared; organizations may restrict third-party systems or use internal models. section 2.2.3
- Use dummy structures and specifications for generated test data when real personal or organizational data would create privacy risks. section 6.2.2 section 6.4
SQL test-data generation
A prompt based on application SQL can generate related rooms and bookings, but the SQL or resulting data must not expose protected user information or restricted intellectual property. Sharing limits determine how the prompt is framed. section 6.2.2
