The Kill Switch: How Corpilus Keeps Your Data Private
The Data Privacy Paradox
AI is most useful when it has access to your most sensitive data — client records, financial reports, strategic plans, HR documents. But this creates a paradox: the same data that makes AI powerful is the data you most need to protect. Every time you send a query to an external AI provider, you are trusting that provider with your company's most valuable information.
Most AI tools handle this with vague privacy policies and data processing agreements. Corpilus takes a fundamentally different approach: the Kill Switch.
What the Kill Switch Does
The Kill Switch is a single toggle in your Corpilus admin panel. When it is on, Corpilus stops sending AI requests to cloud providers such as OpenAI, Anthropic or Google: chat, analysis and document embeddings are processed by a local model, and if no local model is available, the AI request is refused instead of going to the cloud. It covers AI processing, not other integrations you connect.
This is possible because Corpilus can route selected AI workloads to an approved local environment. The exact model and infrastructure choice should be decided during deployment, based on data sensitivity, quality requirements, latency and operational constraints.
Six Layers of Security
The Kill Switch is the most visible security feature, but it sits on top of five additional layers that protect your data at every level:
Layer 1 — Perimeter and application access. The deployment should enforce encrypted communication, restricted access paths and hardened application boundaries appropriate to the customer environment.
Layer 2 — Identity and access control. Users, services and integrations should be authenticated, scoped and monitored according to their role and risk level.
Layer 3 — Tenant and data separation. Customer data must be separated by design, with access limited to the right tenant, workspace and role.
Layer 4 — Secret and key protection. Provider credentials and integration secrets should be protected so they are not casually visible or reusable outside the approved system.
Layer 5 — Data protection policy. Sensitive data should be detected, minimized, anonymized or blocked according to the customer policy before it reaches an AI provider or downstream tool.
Designed around GDPR requirements
For European companies, GDPR compliance is not optional. Corpilus addresses key GDPR requirements: data minimization (the Kill Switch ensures no unnecessary data transfer), purpose limitation (data is used only for the configured AI tasks), storage limitation (document retention policies are configurable), and right to erasure (tenant data can be fully deleted within the system).
Self-Hosted vs Cloud: You Choose
Corpilus gives you a spectrum of privacy options. Some workloads can use cloud AI with protection rules, while sensitive workloads can be routed to stricter or local processing. Most companies need a hybrid policy rather than a one-size-fits-all answer.
The Bottom Line
Data privacy in AI is not about trust — it is about architecture. With the Kill Switch, Corpilus does not ask you to rely on a provider's privacy policy: while the switch is on, AI requests are processed by a local model or refused, and they are not sent to cloud AI providers. That is enforced in the software, not promised in a policy.