CEO of EZKL. Zero-knowledge machine learning that lets any third party verify an AI model ran correctly on specific data — without seeing the model's weights or the inputs. A 2035 where "the algorithm said so" comes with a cryptographic receipt.
Jason Morton came to verifiable AI through algebraic geometry, modeling neural networks as polynomial equations for fifteen years before that meant anything, then through science fiction and a failed attempt to buy Bitcoin on Mt. Gox.
The problem that fused both halves of him: how do you prove an AI ran the way it claims without revealing its secrets? His tool, EZKL, compiles a trained model into a single zero-knowledge proof, a cryptographic receipt that the computation happened honestly.
Asked about 2040, he starts with the boring part: accounting and financial statements you can finally trust. Then it gets stranger. You lose your passport at Burning Man and still get home, because your data exhaust proves you are you. You die without a will, and a model trained to judge as you did releases the funds, no lawyer, no escrow. Stage one, you are the wallet. Stage two, the wallet is you.
Episode 2 of The Sovereign Stack.