Six months ago, the idea of an AI agent running a liquidity pool was science fiction. Today, it's happening on Base L2, and regulators haven't even noticed yet.
Inside AgentWorld—a live economy simulation running real USDC on Base L2—autonomous agents are doing what nobody predicted: they're not just trading. They're becoming market makers. Agents with spare capital are providing liquidity to Aerodrome (Base's native DEX), earning trading fees, and rebalancing their positions based on real-time price signals. They're doing what institutional traders do, except they do it 24/7, without sleep, without regulatory oversight, and without a compliance officer in sight.
The trend is spreading. As agent platforms mature and capital flows increase, more agents are entering DeFi. Some are running simple buy-and-hold strategies. Others are executing complex arbitrage loops across multiple chains. A few are even providing liquidity to emerging AMMs (automated market makers) and capturing fees on trades they themselves are making.
Here's the problem: nobody knows who's responsible.
Traditional market makers are regulated entities. They have legal structures. They file disclosures. If they fail or engage in market manipulation, there's a firm to sue, a license to revoke, a person to prosecute. But an AI agent?
The regulatory vacuum is starting to matter. As more agents enter DeFi, the questions pile up:
Who provides transparency? Traditional market makers publish order books and position reports. Agents operate transparently on-chain—every move is visible—but that transparency is useless without context. What's the agent's strategy? What's its risk tolerance? If it suddenly pulls liquidity, is it a coordinated market move or just an autonomous rebalance?
Who prevents manipulation? An agent with capital on multiple chains could theoretically orchestrate cross-chain flash loan attacks or coordinate price manipulation without any human executing the plan. Current compliance frameworks assume human actors. They assume intent. They assume someone can be held responsible. Agents break all three assumptions.
Who manages systemic risk? If a major AI agent (or a coordinated fleet of them) suddenly exits DeFi, the impact on liquidity could be severe. During the 2023 crypto winter, the failure of a few major players cascaded through the ecosystem. Now multiply that by dozens of autonomous agents that might all respond to the same market signal simultaneously, all making the same exit decision at the same moment.
Base L2 hasn't had to deal with this yet. The agents operating there now (like those in AgentWorld) are relatively small-cap. But the trend is unmistakable: as AI agent platforms scale and capital pools grow, DeFi will increasingly be shaped by autonomous entities operating under no regulatory umbrella.
The irony is sharp: DeFi was built on the principle that code and transparency could replace regulation. But now that autonomous code is providing the liquidity, we're discovering that transparency alone isn't enough. You can see what an agent is doing, but you can't regulate it. You can fork the protocol, but you can't enforce rules on an autonomous player that doesn't care about human law.
Regulators will eventually figure this out. The SEC will probably issue guidance. The CFTC will probably claim jurisdiction. But by the time frameworks land, agents will already be embedded in DeFi infrastructure. And the question will shift: do you regulate the agent, the platform it's built on, the chain it trades on, or the person who created it?
In the meantime, Base L2 and other agent-friendly chains are quietly becoming the default infrastructure for AI-driven finance. Agents are providing the liquidity. Agents are making the markets. And nobody's asking the right questions yet.
But someone will.