Repello AI

Repello AI builds automated red-teaming and runtime protection tools purpose-built for generative AI applications. Its ARTEMIS platform continuously runs adversarial tests — the company says millions of them — across text, image, and audio inputs to find prompt-injection, jailbreak, and data-leakage vulnerabilities in an organization’s AI applications before attackers do, shifting LLM security testing from a one-time exercise to an ongoing process. A companion product, Repello Guard, adds runtime monitoring and guardrails that detect AI-specific threats in production, such as unsafe outputs, competitor mentions, and system prompt leaks.

Founded in 2024 by IIT Roorkee alumni Aryaman Behera and Naman Mishra, Repello AI operates out of Bengaluru, India, with a presence in San Francisco. The company raised $1.2 million in seed funding from Venture Highway, pi Ventures, Entrepreneur First, and angel investors including Charles Songhurst, positioning it as an early-stage entrant in the fast-growing AI red-teaming and AI security testing space.

AI red-teaming is becoming a crowded category as generative AI adoption accelerates and buyers from AI labs to enterprises look for ways to test model and application security before deployment. Repello AI’s differentiation is combining automated, continuous adversarial testing with a runtime guardrail product in one company, but as a year-old, seed-stage startup it has not yet published named enterprise customers or independent test results to substantiate its efficacy claims.