Opsin

Opsin, founded in 2024 by James Pham, Oz Wasserman, and Jeremy Mailen, builds a security layer for enterprise generative AI tool usage, focused on Microsoft Copilot, Google Gemini, and similar assistants. Its platform continuously maps which AI agents exist across an enterprise, assesses what data each agent can surface, and detects oversharing and policy violations in real time — addressing the risk that an AI assistant with broad permissions inadvertently exposes sensitive data a user shouldn’t see.

The company is very early stage: public profiles describe as few as three employees shortly after founding, with a pre-seed round reported at around $3 million in mid-2024, and a separate profile describing a later seed round near $7 million in 2025. Founder James Pham previously worked on machine learning and data security at Abnormal Security, which gives the team relevant domain background even though the company itself has little independent track record yet.

Opsin’s differentiation is narrowly targeting the specific, fast-growing risk of GenAI oversharing inside tools enterprises are already rolling out broadly, rather than general LLM security. Given its very early stage, public evidence of named customers, measurable efficacy, or market traction beyond funding reports is essentially absent, which should temper expectations accordingly.