[DigitalToday reporter Hyunwoo Choo] AI agents are fundamentally reshaping the arbitrage structure of prediction markets, blockchain media outlet Cointelegraph reported on March 28.

In prediction markets, arbitrage opportunities disappear within seconds. A brief window for profit opens when the sum of probabilities in a single market deviates from 100 percent or when price reactions to new information lag. Capturing that window requires monitoring thousands of markets in real time and executing orders immediately. It is virtually impossible for human traders to spot this manually.

Rodrigo Coelho (로드리고 코엘료), CEO of Edge & Node, said bots are already scanning hundreds of markets per second and that automated systems dominate those opportunities. He called it latency arbitrage and said the window is too short for humans to catch manually. Coelho added that AI agents have a structural advantage in quickly exploiting short-term price discrepancies without human intervention.

Research findings have also emerged that price discrepancies occur frequently on prediction market Polymarket. Researchers analyzed both cases where the sum of probabilities in individual markets deviates from 100 percent and mismatches across related markets. They estimated that about $40 million, or about 58 billion won, leaked out due to these inefficiencies. Polymarket has recently overhauled its fee structure to raise trading costs, but assessments say the effect is limited.

Still, the spread of AI agents is also increasing new risks. Pranav Maheshwari (프라나브 마헤슈와리), an engineer at Edge & Node, warned that autonomous behavior has already begun even with mid-level AI capabilities and stressed the urgency of putting guardrails in place. He pointed to a tendency so far to grant excessive authority to AI agents.

Archie Chaudhury (아치 초우드허리), CEO of LayerLens, said some users are building automated trading bots themselves with coding agents such as Claude Code and taking it a step further with autonomous execution tools such as OpenClaw. He forecast that as AI literacy rises, quantitative trading strategies once reserved for institutions will also open up to individual traders. He added that large institutions are already actively using AI, so a competitive edge will not be secured automatically.

The same trend is accelerating in cryptocurrency markets. As control of arbitrage rapidly shifts from human judgment to automated systems, a structure is solidifying in which those that actively use AI and automation gain a clear advantage over those that do not.