Not used for model training
Prompts you submit, documents you upload, model outputs, and retrieved business information are used to process the task you start. They are not used to train models.
Collective intelligence for clearer strategyLight Your Path. Ever By Your Side.
Your business information, carefully protected.
We choose model services that do not use prompts, uploaded documents, model outputs, or retrieved information for training, and protect this content through data isolation, encryption, and access controls.
Three core protections
Customer information is isolated
Customer information is managed within isolation boundaries and protected by encryption in storage and in transit. Retrieval vectors are also treated as sensitive data and included in isolation, encryption, and deletion controls.
Access requires your authorization
Permissions, access records, and monitoring for unusual activity govern data access. We access relevant data only within the scope you authorize, to address issues you report or improve service quality.
How tasks use information
Context is selected for each task
Lumos selects the documents and information the current analysis needs. It does not indiscriminately send the entire business knowledge base to the model.
Execution stays bounded
Tasks that require execution run in restricted environments. Tool, file, and network permissions are set according to each task. Permission controls, environment isolation, and related measures reduce risks from malicious instructions in documents or web pages.
Lumos provides analysis support. You and your team make the final decisions. Legal and other high-risk conclusions still require review by relevant professionals.
Production changes are reviewed
Production changes go through code review, approval, and controlled release. Infrastructure rules are managed as code. Before changes, the relevant modules, call relationships, and key security constraints are identified to assess their security impact.
How information is processed and retained
Model providers process data needed for inference
Lumos supplies relevant context for each task. Model providers process the data needed to complete that inference request.
Zero data retention can be confirmed for sensitive work
Lumos has a zero data retention arrangement for highly sensitive work that can constrain a model provider’s retention of data from a single inference request. It is not enabled by default for every request. The provider, features, data scope, and restrictions on human review must be confirmed before engagement. Zero data retention does not cover materials, memory, and results you choose to save in Lumos.
Temporary and saved information are managed separately
Task execution environments are destroyed when tasks end. Temporary context and caches have shorter lifecycles. Materials, memory, and results you choose to save remain for later work and are not cleared when a task ends. The scope and timing of an active deletion, including retrieval vectors and backups, need separate confirmation.
Frequently asked questions
Will my information be used to train models?
The prompts you submit, documents you upload, model outputs, and retrieved business information are used to process the task you start. They are not used to train models.
Will third-party models process my information?
Model providers process the data needed for inference. Lumos selects relevant context for each task and does not indiscriminately send the entire business knowledge base to the model.
A zero data retention arrangement is available for highly sensitive work. It is not enabled by default for every request. The provider, features, data scope, and restrictions on human review must be confirmed before engagement. It does not cover materials, memory, and results you choose to save in Lumos.
Is everything deleted when a task ends?
No. Task execution environments are destroyed when tasks end, and temporary context and caches have shorter lifecycles. Materials, memory, and results you choose to save remain available for later work.
Provider zero data retention does not mean Lumos does not store these materials. The scope and timing of an active deletion, including retrieval vectors and backups, need separate confirmation.
Can we use our own keys?
Using your own keys can be discussed as a custom solution. Delivery scope must be confirmed before engagement. It is not included by default in every plan.
Can you guarantee that information will never be exposed?
Lumos uses customer information isolation, encryption in storage and in transit, role-based authorization, access records, and monitoring for unusual activity. These measures reduce risk. They do not guarantee that exposure can never occur.
We access relevant data only within the scope you authorize, to address issues you report or improve service quality.
Read our privacy policy for how personal information is processed, your rights, and how to contact us. For work involving highly sensitive materials, confirm the processing scope and applicable conditions first.
Read the privacy policy