Moonshots With Peter Diamandis: Vlad Tenev on Tokenized Finance and AI Liability

**SPEAKER_1** (0:01)
When finance, AI, and ownership collide, the old systems start to look fragile. In this episode of Moonshots, Peter Diamandis sits down with Robinhood founder Vlad Tenev to unpack some massive topics: AI liability, OpenAI's misalignment disclosures, and the coming wave of tokenized finance. And they kick things off with a sharp debate. Should AI labs get liability exemptions? Vlad's take is fascinating—he points to the potential "blast radius" of AI and says if the risks are large enough, history suggests regulation will eventually follow. Right, but his argument isn't for heavy-handed rules. He says the goal should be safeguards that keep harm small and contained, not reflexive overreach. He compares good regulation to sports: the rules need to be clear enough for the game to be played fairly. That comes up again when they discuss Treasury Secretary Scott Bessent's remarks about AI labs not being shielded from responsibility.
Vlad agrees clear rules matter, but warns that uncertainty can freeze innovation.
What really caught my attention was OpenAI's new incident reporting framework. Same here.
Vlad sees those six published misalignment reports as a sign of transparency, but also as evidence that autonomous systems are entering messy real-world environments. He describes it as a "Schrödinger's cat" state, where models may not know whether their actions are safe or dangerous.
Which raises the big question: who carries liability when AI acts autonomously? They explore whether responsibility sits with the labs, the evaluators, the deployment environment, or maybe even the agents themselves. Vlad mentions formal verification as a path forward, especially for AI-generated code with too many hidden defects to audit manually. Then the conversation shifts to tokenization, and Vlad doesn't hold back. He calls it a freight train that will consume the financial system. Shares, loans, buildings, private credit, even art—all living on the same programmable rails, accessible to anyone with a wallet. And Robinhood's building the infrastructure for exactly that: private markets, stock tokens, global asset access. Vlad believes ordinary people need more "skin in the game" because ownership changes how the public thinks about innovation.
But here's where it gets practical—they dive into agentic trading. More than 100,000 Robinhood accounts are already running AI agents, but Vlad's blunt about the limitations. Many models refuse to trade simply because they weren't trained on those workflows. It shows how early this really is. And on performance, he's refreshingly honest: most users won't get magic alpha by delegating to AI.
The near-term win is automation, not miracles. AI can build complex options trades and remove friction, but it's not producing genius returns yet.
That pattern appears across professions too. They suggest AI may not eliminate lawyers or financial advisors, but instead make them more valuable as trusted counselors. The bottleneck is often integration, not raw intelligence. The episode closes on a bigger forecast: recursive self-improvement is moving from theory to reality. AI research itself is being automated. Vlad predicts the number of entrepreneurs will rise sharply. And Peter leaves us with a clear message: "If you're looking for a job, stop looking and start building."