Skip to content
Companies

Faberlens

Behavioral risk testing and runtime policies for AI agents

Raised to date

$375K

of $750K offered across 2 Form D filings

Still open
$375K
Last filed
Apr 27, 2026

About

Faberlens measures how an AI agent behaves as a whole, meaning its model, harness, skills and MCP connectors together, and then governs that behavior. Its Prove It product runs probes covering normal use and attacks against the full agent setup and reports what regressed. Its Protect It product discovers the policies a given setup needs and enforces them at runtime. The company is onboarding launch partners and takes demo requests.

Tests the specific combination of model, harness, skills and connectors rather than the model alone, and produces evidence for ISO 42001, NIST AI RMF and EU AI Act reviews.

Funding

  • OtherFiled Apr 27, 2026Form D/A
    $375K/ $750K
  • OtherFiled Apr 13, 2026Form D
    None sold/ $750K

Product

Prove It

AI agent evaluation

Behavioral risk measurement for a specific agent composition, using probes derived from that composition and run against the base model and the full setup under the same conditions.

  • Probes for normal operation and adversarial attacks
  • Scorecard comparing base model and full composition
  • Continuous re-evaluation as the agent setup drifts

Built for

Enterprises deploying agents such as Claude Code, Cursor or Codex with MCP connectors and skills, and teams building custom agents

People on file