Company

Gemisis is a Canadian technology venture providing traceability for internal and external communications. Outgoing communications are logged against exact linked senders. Receivers independently compare specific communications with those sender-side records.

Current coverage includes Gmail, Outlook, Meet, Teams, Zoom, phone calls, AI agents, wire-transfer recipients, domains, and ecommerce. Available checks depend on the integration, permissions, and evidence.

Silicon Valley is caught in a massive paradox: investing billions to make AI agents indistinguishable from humans, while simultaneously investing in 'liveness detectors' to block them. Detection is a losing battle. The future isn't about blocking AI agents; it is about authenticating their root of authority.

Gemisis focuses on the sender behind a specific communication. The receiver independently compares it with registered outgoing evidence, rather than guessing whether the interaction looks human.

I spent the last several months inside the finance department of a company that does roughly $50 billion in revenue a year - the one that owns Ray-Ban - working on reconciliation: making sure both sides of a transaction were accurately who they claimed to be, at a scale where even a tiny gap opens the door to internal fraud. That's the same underlying pattern as Gemisis, just pointed outward instead of inward - verifying that a claim of identity actually matches reality, before money or trust changes hands.

At the same time, I've watched AI make that kind of deception easier every month, not harder. It's not a future risk anymore. It's already happening to institutions and to individual families, and there's no reason to think it slows down.

I care about this idea specifically, not just about starting a company. I think about it daily, and I think it points toward a future I actually want to help build - one where people can trust who's really reaching out, no matter how good AI gets at faking it.

Gemisis starts with institutions - banks, insurers, government agencies - because that's where impersonation does the most immediate, quantifiable damage. But the underlying problem isn't limited to institutions. Voice cloning has already been used to convincingly fake a family member in distress, asking for help, on a call that sounds exactly right. As that gets easier, I believe people will want the same kind of verification for each other that institutions need today - a real, checkable answer to "is this actually them."

That longer-term vision builds on the same approach: link the sender, record the outgoing communication, and let the receiver independently compare the evidence.

Use the official Gemisis contact page to reach the team. Security researchers should review the responsible disclosure guidance before reporting a potential vulnerability.