Retail-brokerage
Public launches AI agents that trade prediction markets through Kalshi, and link the odds straight into stock and bond positions
The agentic brokerage's new feature lets members instruct an AI agent to watch event probabilities on crypto, commodities, climate, economics and politics, then trigger trades automatically, with Kalshi supplying the CFTC-regulated exchange rail underneath.
Public built its identity as 'the world's first agentic brokerage', so this is a natural extension rather than a pivot, but the mechanism is worth reading carefully. Members either trade prediction-market events directly or hand an agent a standing instruction, and the agent turns that into fixed rules: send an alert when odds move past a threshold, or place an order automatically when they do. Execution and market infrastructure run through Kalshi, which the announcement describes as the largest prediction market and a CFTC-regulated exchange, so the regulatory perimeter here is the same one this site has tracked around Kalshi all year, not a new one Public is opening on its own.
The more interesting design choice is treating event odds as a signal that feeds conventional stock and bond trades, not just a separate betting product bolted onto a brokerage app. Co-CEO Leif Abraham's framing, that prediction markets give agents 'a new input: real-world events', is the tell: the pitch is an agent that watches a Fed-decision or election-outcome market and repositions an equity or bond holding off the probability shift, collapsing what used to be two separate trades (bet on the event, then separately trade the correlated asset) into one automated instruction. That is a genuinely new agent workflow, distinct from the order-routing and portfolio-monitoring agents this site has covered from Robinhood, eToro and Coinbase.
It also sharpens a question regulators have been circling without resolving: an agent that auto-trades a regulated security off an unregulated-adjacent signal (event odds) blurs a line that existing agentic-trading oversight, built around order execution and account permissions, was not obviously drawn to catch. Public's structure keeps a human setting the rule and the cap, the same conventional guardrail every other agentic brokerage still leans on, so the product itself is not reckless. But it is one more instance of an agent's inputs (external event probabilities) sitting one step removed from the trade the agent actually places, which is exactly the kind of composed automation the FSB and IOSCO papers this site has tracked keep flagging as harder to supervise than a single-model trading bot.
Read the original: PR Newswire - Public launches AI agents that trade prediction markets through Kalshi, and link the odds straight into stock and bond positions. Commentary is the independent editorial view of Share Trading; the original article is credited to its publisher.