About Brava

Making Detection Provable

Security can score 100% on its own tests - and still fail to recognize critical attacks. As security operations become autonomous, AI can turn incomplete evidence into false certainty - then act on it at machine speed.

OUR STORY

Why We Built Brava

The problem is circular: organizations test against attacks they already know to look for, then mistake a passing score for proof of protection. We founded Brava to break that circle.

Our cooperating AI agents start with what attackers could do and synthesize the digital evidence those attacks would leave across each organization's systems - independently of existing detection rules or predefined tests.

That independent ground truth allows enterprises to establish what their defenses should recognize, verify what they actually detect, repair failures, and prove the fixes work.

Our ambition is to build the machine-readable ground truth for threat detection - both for humans and AI agents.

THE FOUNDERS

Idan Basre

CO-FOUNDER & CEO · New York

Seventeen years seeing cybersecurity from both sides. A Talpiot alumnus, Idan spent a decade researching attacker behavior in Unit 8200 before moving into product leadership at Microsoft Defender. He brings together the attacker's imagination and firsthand experience building defenses at global scale - a combination that shaped Brava's founding question: Would your defenses recognize the attack?

Itay Gabbay

CO-FOUNDER & CTO · Tel Aviv

Fourteen years leading high-velocity engineering organizations, including as VP R&D. Itay led joint US–Israeli DARPA AI research, developed autonomous AI systems, and led enterprise AI security initiatives. His fundamental question at Brava: What if every security system could learn and improve from the same evolving understanding of attackers?

THE TEAM

BACKED BY

Every Day, at Customer Scale

What Brava processes in a single day across customer environments. Even at that volume, the evidence of a critical attack can be missing. Brava establishes what should be there.

2.4 trillion

events read and fingerprinted a day

1,100

attack techniques replayed a day

3,800

detections re-tested against that evidence

140 K

fields mapped, including the ones no rule reads

Rounded, measured over a 24-hour window across customer environments.

Join the Team

We are growing the team. Send us a note about what you would like to work on.

See Open Roles Book a Demo