depthfirst Raises $80M Series B to Train Security Models That Understand Real Exploitability
Security used to be a checklist. Now it’s a chess match where the board keeps changing mid-game. depthfirst just pulled in $80M in Series B to operate in that chaos, right where most tools start guessing and attackers start grinning.
San Francisco keeps producing these quiet monsters. depthfirst, founded in Oct. 2024, is not playing the volume game. It is playing the accuracy game. Co-founder Daniele Perito, who has already seen what happens when security meets scale at Cash App and Faire, teamed up with Co-founder and CTO Andrea Michi, fresh out of Google DeepMind’s reinforcement learning trenches. That pairing tells you everything. One has lived the pain, the other knows how to teach machines to hunt it.
Meritech Capital led the round, with Forerunner Ventures and The House Fund stepping in, and Accel, BoxGroup, Liquid 2 Ventures, Alt Capital, and Mantis VC doubling down. Smart money does not chase noise. It chases signal. And right now, signal says security needs to grow up.
depthfirst’s angle is simple to say, hard to build. Their General Security Intelligence platform does not just look at code snippets and throw alerts like confetti. It builds context across code, infrastructure, and business logic, then asks the only question that matters. Can this actually be exploited in the real world? If the answer is yes, it does not just panic. It hands developers something they can merge.
Then comes dfs-mini1, their in-house model, trained with reinforcement learning in security-specific environments. Not a generalist trying to moonlight as a security guard. A specialist. On OpenAI’s EVMBench for smart contracts, it is putting up state of the art performance while running at a fraction of the cost. Cheaper, sharper, and less distracted. That combination tends to win.
The timing is not accidental. Software is being written faster than it can be understood, and attackers are now carrying the same AI tools as builders. That arms race does not reward teams that rely on yesterday’s filters. It rewards teams that can see systems the way attackers do, end to end, pressure points included.
This raise brings depthfirst to $120M total in just a few months. That kind of acceleration usually means one thing. The market is not politely interested. It is a little uneasy, and it should be. Congratulations to Daniele Perito, Andrea Michi, and the entire depthfirst team. Also a nod to the investors who recognized that in a world full of noise, the real edge is knowing exactly where things break before someone else does.
