We watched the data walk out —
so we built a watchdog.

In the winter of 2025, Maya Ellison was leading a security review at a New York financial firm when she found three months of source code sitting in a consumer chatbot's history. Nothing had been hacked. Nobody had broken a rule on purpose. The data had simply walked out through the front door — pasted, one helpful prompt at a time, into tools the company never approved and couldn't see.

She called Daniel Okafor, an infrastructure engineer she had trusted since their first incident response together. Two weeks later, in a cramped Flatiron sublet, they had a working prototype: a proxy that could see AI-bound traffic, classify what was in it, and stop the sensitive parts before they ever left the building.

That prototype became Watchdog AI. In 2025 we raised a seed round from enterprise-security investors, signed our first design partners — three New York fintech security teams — and spent the year learning exactly how data leaks into models, one endpoint at a time. In early 2026 the platform reached general availability, and today teams across finance, healthcare and technology run Watchdog between their people and the models they want to use.

Our belief is simple: companies shouldn't have to choose between moving fast with AI and keeping control of their data. The watchdog stays on duty so your team doesn't have to think about it.