Refractal is an AI security company building infrastructure to protect AI agents and LLMs from prompt injection, jailbreaks and data exfiltration.
Refractal is an AI security company building the security infrastructure organisations need to defend against autonomous AI threats. Its work centres on protecting AI agents and LLMs from prompt injection, jailbreaks and data exfiltration - problems where failure is not an abstraction. The company serves governments and organisations that cannot afford to get security wrong.
Refractal's platform enables organisations to automatically stress test their AI deployments, safely evaluate AI cyber risk, and enforce policy across models, tools, data and modalities. The technical scope is broad and demanding: prompt injection defense, jailbreak prevention, data exfiltration prevention, AI agent security, and AI red teaming and stress testing, combining proven cybersecurity methods with frontier AI security research.
The team's expertise is drawn from MIT, Microsoft, NASA and the U.S. Navy. Refractal's approach to AI risk is deliberately practical, and its work is aligned with emerging standards including the EU AI Act, NIST AI RMF, ISO/IEC 42001 and SOC 2 Type II. The company is backed by leading deep-tech and cybersecurity investors.
