Which data?
Identify documents, personal data and confidential content. Clarify the purpose, legal basis and permitted use.
DATA PROTECTION IN AI PROJECTS
Confidential data.
Under your control.
With Mikis, you can also use AI for confidential data. Local language models process your inputs and internal content within your controlled infrastructure. In this deployment, content does not need to be sent to external AI providers.
These questions belong in project planning.
Which processing?
Review imports, vector data, model requests, answers, logs and backups. Explicitly identify external services.
Which responsibilities?
Agree on access rights, deletion, support, contracts and required safeguards with the responsible parties.
From objectives to a configuration you can assess.
For fully internal operation, document preparation, embeddings and retrieval are configured within the controlled environment alongside the language model. Your documents serve as a knowledge source; no external model training is required. Access, logs, backups and deletion are defined to meet your protection requirements.
See the confidential business communications case study for a practical example. Optional external models and interfaces receive only the data approved for the relevant application. Before the pilot, we clarify suitable test data and any required data protection assessments with the responsible parties.
LET'S TALK
Clarify data and requirements
together.
What data would you like to use? What requirements does your organisation have? Together, we discuss the right next step.