90 Days Gen AI Risk Trial -Start Now
Book a demo
AI Model Construction·epsilla.com

Epsilla

Epsilla enables the creation of vertically-focused LLM agents through reinforcement learning, facilitating AI model construction with customization using private data and knowledge.

Risk Score
High
6/10

Independent assessment across data handling, compliance, security and transparency.

Risk factors

3
  • Uses private data for model customization
  • Potentially trains on user data without clear opt-out
  • Requires access to sensitive data for effective model training

Recommendations

5
  • Request SOC 2 and tenancy isolation details
  • Execute DPA covering training data
  • Scope data ingestion to non-sensitive sources
  • Confirm encryption and key management
  • Define customer-data training boundaries contractually

Data handling

Storage
Not publicly documented
Retention
Not publicly documented
Training on inputs
Customer data used to fine-tune/RL agents per product description; boundaries unclear