data-exfiltration-prevention
(3)
Silmaril
Silmaril is a San Francisco security startup building a runtime firewall for AI agents and AI-native applications. It intercepts prompt injections, context poisoning, and dangerous tool calls in real time - a self-healing classifier that hunts for new attacks against a customer's own environment and retrains itself continuously. Founded in 2026 by Aum Upadhyay (ex-AWS security) and Eduardo Velasco (whitehat who has found exploits in major AI systems), Silmaril is part of Y Combinator's 2026 batch.
ORION Security
ORION Security is an AI-native data loss prevention (DLP) company that protects enterprises from data leaks without relying on manually written policies. Founded in 2024 by Nitay Milner and Jonathan Kreiner, ORION uses large language models and specialized AI agents to map how data normally flows across SaaS, email, cloud, endpoints, and AI tools, then analyzes content sensitivity, user identity, behavioral intent, and data lineage to catch exfiltration and insider threats in real time. The platform targets the three core sources of data loss - human error, malicious insiders, and external attackers - while cutting the false positives and maintenance burden that plague legacy DLP.
Cyberhaven
Cyberhaven is a data security company that protects enterprises from data exfiltration and insider risk by tracing how information moves rather than just where it sits. Founded in 2016 by five PhD researchers with roots in the DARPA Cyber Grand Challenge, the company built a proprietary 'data lineage' engine and an AI layer, Linea AI, that combines Data Loss Prevention (DLP), Data Security Posture Management (DSPM), Insider Risk Management, and AI security into one platform. Reaching a $1 billion valuation in April 2025 after a $100M Series D, Cyberhaven serves customers such as Snowflake, Motorola, Reddit, and major law firms.