Joseph Siffred Duplessis, Benjamin Franklin, 1778
Ollive AI Risk Labs studies how AI agents fail, why they lead to losses, and how to make AI technology insurable.
In 1894, fire insurers created Underwriters Laboratories because no one could tell a safe electrical product from a fire waiting to happen.
Automobiles had crash testing. Aviation had certification. Each technology became trusted because someone built an independent way to measure and underwrite risk.
AI agents are that technology now and Ollive AI Risk Labs is building that institution.
Benchmarks tell you what an AI model can do.
They don't tell you what happens when it's wrong.
We study the failures that matter in production—from hallucinations and privacy leaks to financial loss, regulatory exposure and legal liability—because those determine whether enterprises can trust AI.
Every AI failure follows the same chain:
Failure → Harm → Liability → Financial Loss
AI Risk Labs turns that chain into evidence through the Agent Trust Score—an independent measure of how safe is it to deploy the AI agent.
Not another leaderboard. A trust signal for the real world.
Anyone can publish a benchmark.
Our research directly informs insurance coverage for AI companies.
When our understanding of risk improves, customers benefit. When we're wrong, reality corrects us through claims and loss data.
That's how safety institutions have improved every major technology—and how AI should too.
The most important AI failures are not visible in benchmarks. They happen in production
Ollive AI Risk Labs combines real-world deployments, adversarial evaluations and insurance data to build the risk infrastructure for trustworthy AI.