01
Data isolation
Tenant- and deal-level separation of every workspace. Your documents, embeddings, and agent outputs are isolated per client — with row level security enforced at the user, server, and database levels.
Security
Your data stays yours.
Enterprise-grade security for high-stakes financial work. Midlevel is built for firms who want to keep proprietary knowledge under their control.
02
Model privacy
Confidential deal documents don't belong in generic LLMs. Our model-agnostic harness lets you use the models you prefer while keeping source documents in a governed, encrypted environment, outside of LLM memory. Your prompts, documents, and outputs are never used to train shared models.
03
Access control
MFA, SSO, and role-based permissions controllable by your workspace admin. Immutable activity logs across every workspace.
04
Encryption
AES-256 at rest, TLS 1.2+ in transit. The same level of encryption used by financial institutions.
05
Compliance
- NDA compliant archiving and retention
- PITR disaster recovery through Cloudflare and AWS
- GDPR and CCPA compliant
- ISO and SOC 2 Compliance — audit in progress, anticipated January 2027
- Sub-processor list: published in privacy policy
- Retention & deletion: client data deleted within 30 days of termination
- Annual third-party penetration tests
For additional information contact info@midlevel.ai