Nitay Milner\u2019s entrepreneurial story starts long before cybersecurity. At 13, he took the money from his bar mitzvah and turned it into a DJ business. By 18, he already had five employees and real revenue. \u201cI was always that kid who needed something big, something interesting, something that moves the needle,\u201d he told me. That itch\u2014for ownership, for building\u2014never left.<\/p>\n\n\n\n
After serving in the financial advisor to the Chief of Staff unit in the IDF, helping manage tens of billions in defense budgets, Milner studied computer science and deliberately chose product management as his path into tech. \u201cProduct felt like a mini-CEO role,\u201d he said. \u201cBalancing customers, R&D, and market fit\u2014that\u2019s the best preparation for founding.\u201d At Epsagon, and later at Cisco Cloud Observability after the $500M acquisition, he led cloud-native and observability products before making the jump into security entrepreneurship.<\/p>\n\n\n\n
Orion Security was born from a deliberate, methodical ideation process\u2014not a midnight epiphany. Milner and his co-founder, Jonathan, ran a Converge\u2013Diverge process: forty ideas in a Notion table, many intentionally \u201cstupid,\u201d then weeks of pruning based on feasibility, differentiation, and passion. They focused on security because both had lived the pain\u2014Milner from the product side at Cisco, Jonathan from years in Unit 8200\u2019s offensive ranks. After dozens of conversations with CISOs, one theme kept surfacing: everyone hated DLP, but everyone needed data protection.<\/p>\n\n\n\n
\u201cHalf the CISOs told us, \u2018You look young and ambitious\u2014don\u2019t waste your best years on DLP,\u2019\u201d Milner laughed. \u201cBut when we asked about their biggest security problems, they always came back to: \u2018I just don\u2019t want sensitive data leaking out.\u2019 They wouldn\u2019t say \u2018DLP,\u2019 but they were describing DLP.\u201d<\/p>\n\n\n\n
The challenge was clear: the category<\/em> was right, the implementation<\/em> was broken. Traditional DLP depends on brittle, static policies and mountains of manual work. Deployment cycles run 8\u201312 months. False positives flood teams. Only the largest enterprises can afford to staff it. Milner and his co-founder sat in a room and asked themselves a blunt question: Are we just building \u201canother DLP,\u201d or can we design something fundamentally different\u2014something that uses LLMs and intent-based analysis to make DLP 100x better at a tenth of the operational cost?<\/p>\n\n\n\n Their thesis: legacy DLP is anomaly detection for people, and people are inherently anomalous. Machine learning on behavior alone was never going to be enough. But LLMs bring something qualitatively new\u2014contextual understanding. Orion\u2019s approach is to treat every potential data exfiltration event as if a smart analyst were sitting there, looking at the user, the content, the destination, and the business context, then deciding: is this actually risky or just work-as-usual? \u201cThink of it as moving from \u2018did something weird happen?\u2019 to \u2018did something dangerous happen?\u2019\u201d Milner explained.<\/p>\n\n\n\n