Depthfirst Secures $40M in Series A Funding for Business Security Solutions

Depthfirst showed up in San Francisco in October 2024 with a very specific irritation and zero interest in pretending the industry was fine. Software is now written faster than humans can reason about it, thanks to AI copilots shipping code like a sleepless intern on cold brew. Legacy security tools are still scanning like it is 2012, matching patterns, flooding teams with noise, and calling that protection. Depthfirst looked at that gap and decided to build something that actually thinks.

General Security Intelligence is not a tool watching from the sidelines. It operates like a senior security engineer who understands the entire system, remembers context across cycles, and does not check out at 3 a.m. Instead of heuristics, it learns. Instead of alert storms, it reasons. Code, dependencies, containers, infrastructure, cloud. Full stack, full context, no excuses. That distinction matters when AI-generated code is pouring into production faster than any human team can review it.

That philosophy just pulled in a $40M Series A led by Accel, with Alt Capital, BoxGroup, Liquid 2 Ventures, Mantis VC, and SV Angel leaning in. When Jeff Dean and Kirsten Green show up as angels, that is not curiosity, that is conviction. The round landed roughly four months after general availability, which says a lot about how fast real security teams recognized substance instead of theater.

The founding lineup explains the pace. Qasim Mithani scaled infrastructure at Databricks and AWS and knows exactly where systems fracture under pressure. Daniele Perito built and defended complex marketplaces at Square and Faire and understands how risk actually propagates. Andrea Michi brought DeepMind-grade reinforcement learning into production reality. Jeeyoung Kim scaled critical systems at Plaid and Square with zero tolerance for fragility. This is not résumé stacking. It is lived experience colliding in the right order.

The outcomes speak clearly. 8x more true positives. 85% fewer false alarms. A 53% success rate on CyberGym where the historical baseline sits around 20 to 28%. The discovery of CVE-2025-59419 in Netty, a library embedded across Netflix, Apple, Amazon, and more, was not a marketing flex. It was proof that contextual understanding beats signature guessing every time.

Customers like AngelList, Supabase, Moveworks, and Lovable are not buying theory. They are buying leverage. In a $400B market where 92% of organizations plan to increase security spend by ~27%, Depthfirst is not trying to be louder. It is going deeper, first, and letting the system do the talking.