Kalshi's AI Agent Scrubs Prediction Market Contract Language

Kalshi built an artificial intelligence agent to vet the wording of its prediction market contracts, co-founder Luana Lopes Lara said in a Bloomberg interview Monday.

The tool is meant to catch phrasing problems before a contract reaches the floor. That matters because small wording differences have already caused controversy. Lopes Lara pointed to a January bet on whether a Netflix executive would say "Warner Bros." during an earnings call. The executive said "Warner Brothers." Kalshi ruled the answer was no.

The AI agent does more than check contract language. It aggregates top news, scans competitor listings, and recommends new contracts for Kalshi to launch, Lopes Lara said.

"We actually have an AI engineer in the markets team, where the AI is battle-testing the entire certification – finding out if you go in this direction, maybe there's a hole here, and all of that," she said.

The company earlier this month was also building a new interface for its most active retail traders, according to a CNBC report. The dashboard would let users track popular contracts by 24-hour volume, watch trades as they happen, monitor order books for individual contracts, and view event contracts tied to their own portfolios. CNBC, citing unnamed sources, said Kalshi could eventually add research data and extend the tool to other asset classes. Kalshi did not comment when PYMNTS asked.

The agent news lands as financial regulators take a harder look at agentic AI. The Financial Stability Board earlier this month recommended that financial institutions build safeguards against autonomous agents, warning that risks can develop faster than human oversight can catch.

PYMNTS Intelligence research found 59% of firms face bot-driven fraud as an active threat, and companies lose 3.1% of annual revenue to identity gaps. A core security problem: a human user logs in once and leaves a trail that fraud detection can follow. Agents act continuously, spawn other agents, and carry permissions that spread in ways no one could predict when access was first granted.

Kalshi's AI agent lives inside the markets team, not on the trading side. That limits exposure for now. The same infrastructure that reviews contracts today could eventually touch order flow, data feeds, or listing decisions – places where an agent's permissions would carry more weight.

Lopes Lara did not say when the agent would go live or whether Kalshi would disclose its use to traders who enter contracts it helped certify.