Artificial intelligence security startup Lasso Security Ltd. today launched a guardrail for AI applications that runs on ordinary server processors, with no graphics chip required.
The product is called LEAP. Lasso also announced $30 million in new funding.
Enterprises that put guardrails in front of production AI systems typically run every request through a second large language model, and that model needs accelerated hardware behind it whether the request turns out to be dangerous or not. Lasso said the cost pushes most enterprises into inspecting a portion of their AI traffic and accepting the risk on the rest.
LEAP is transformer-free and returns a verdict in under five milliseconds, the company said, at thousands of times the throughput of existing guardrails. The benchmarks are Lasso’s own.
Co-founder and Chief Executive Elad Schulman said the question of whether enterprises would put AI in front of their customers is “settled,” and that what they now need is security able to inspect not only what agents say but what they do.
Ophir Dror, co-founder and chief product officer, put the objection to the prevailing approach more bluntly. “The concept of a guardian agent or agents overseeing other agents is great, unless you are the one paying the bill,” he said. “In real-world deployments with millions of prompts and actions, this idea is just not feasible.”
LEAP is one of two engines on the platform. The second, RAPID, runs a self-hosted model that rules on the small share of decisions requiring a judgment call against a policy written in plain language, an approach Lasso said costs a small fraction of what the same request would cost through a commercial model provider. A routing layer decides which engine sees what, so most traffic never touches a graphics processing unit.
Lasso came at the problem from the attacking side first. Its offensive team runs automated red teaming against customer applications and agents, working up from single prompts to autonomous multiturn attacks, and what it finds goes back into the product. During one engagement at a global healthcare provider, Lasso said the team pulled every patient record and was able to alter prescriptions.
The software is deployed inside customer environments across financial services, insurance, hospitality and automotive, and Lasso said it now covers tens of thousands of agents and billions of requests and actions a month. Revenue has grown more than 500% in the past 12 months, according to the company, which didn’t reveal an absolute number. BMW AG is a customer, and so are eToro Group Ltd., Fiverr International Ltd. and the U.S. Department of Homeland Security.
ClearSky Advisors led the funding round. Entrée Capital, which led Lasso’s $6 million seed round in 2023, increased its position, and iAngels, Singtel Innov8 Pte. Ltd., Mindset and Swish Data joined.
Schulman and Dror founded Lasso in 2023. The round pays for more engineers across its Tel Aviv and New York offices, the company said, along with a push into federal and regulated-industry accounts and wider sales coverage in North America and Europe. Investors have now put more than $37 million into the company.