Media

Most organizations buy AI the way they buy software. That is the mistake. It is not deterministic, it changes as it learns, and it cannot govern itself. It has to be raised instead: supervised, recorded, and granted authority in graded steps as it earns them.

Electrification abstracted energy. Telecommunications abstracted distance. Digitization abstracted memory. AI abstracts time, and specifically one window: the gap between the moment a consequence becomes visible and the moment it hardens into something you can no longer prevent. Whoever sees more of those windows wins.

You elect a president into an existing constitution. Defined powers, defined limits, things that have to pass through Congress. You do not appoint someone first and work out their authority afterwards. That is exactly what we are doing with AI agents, deploying them into organizations that have never defined what they are allowed to decide.

NHTSA has written to Robotaxis companies about vehicles interfering with firefighters and police at emergency scenes. These are no longer edge cases. The capability arrived faster than the rules governing it, and regulators will not close that gap in time. So who authorizes an autonomous vehicle to enter a crime scene, and who is accountable when it gets that wrong?

At MACHINA 2026, Tal Cohen participated in panel, "Who's Funding the Robot Revolution?" A conversation on where the capital behind physical AI and robotics is actually coming from.

A conversation with Tim Albertsen, former CEO of Ayvens, on what the physical world actually needs from AI, and why the next frontier is not a smarter model, but a smarter habitat.