Goldman Sachs and Morgan Stanley updated their internal compliance policies on March 18, 2025. Employees can no longer trade on Polymarket or Kalshi. The directive cites insider trading concerns. This is not a rumor. It is a documented shift in the regulatory geometry of prediction markets.
Context: The Two Roads to Price Discovery
Prediction markets allow users to trade contracts on future events—elections, product launches, macroeconomic data. Polymarket operates on Polygon, using UMA as an oracle for dispute resolution. No KYC. Pseudonymous wallets. Kalshi, by contrast, is a CFTC-regulated designated contract market. Full identity verification. Legal in all 50 states. Both platforms serve the same function: aggregating disparate information into a market-clearing probability. The difference is the trust model. Zero trust is not a policy; it is a geometry. Polymarket’s geometry is permissionless but porous. Kalshi’s is permissioned but surveilled.
Wall Street’s move targets the porous geometry. The banks did not issue a vague warning. They specified platforms. They named the risk. This is the first time traditional finance has explicitly mapped insider trading concerns onto prediction markets. The code does not lie, but it often omits. What the banks omitted is the actual evidence—no public case of insider trading on Polymarket has been confirmed. Yet they acted. The precautionary principle in compliance is a blunt instrument.
Core: Deconstructing the Regulatory Vector
From a technical standpoint, the banks’ action reveals something important about the information asymmetry surface area. Prediction markets are not casino. They are forward-looking derivatives. When an employee of a pharmaceutical company knows a drug trial result before publication, they can short the “approved” contract on Polymarket before the official press release. The on-chain footprint is permanent. But the link between a wallet and a real-world identity is opaque—unless the employer has deployed blockchain surveillance tools. Banks have access to Chainalysis and Elliptic. They can trace suspicious wallet clusters to internal payroll data. The fear is not hypothetical. It is audit-ready.
Compiling the truth from fragmented logs: in my 2017 audit of the 2x2x4 protocol, I found a reentrancy vulnerability that allowed infinite borrowing. The root cause was not a bug but an unchecked assumption—that the call to the external contract would not recur. Here, the unchecked assumption is that pseudonymity protects insiders. It does not. On-chain analysis has matured. The banks’ internal compliance logs now include transaction histories from DeFi protocols. The geometry of trust has shifted from “who you claim to be” to “what your wallet has touched.”
The Technical Vulnerability: Oracle as Insider Vector
Polymarket uses UMA’s DVM for dispute resolution. UMA token holders vote on the outcome of contested markets. If an insider can influence or bribe UMA voters, they can alter the settlement of a contract—effectively locking in a profit from their private knowledge. This is not a theoretical edge case. In 2022, I assessed the slashing ambiguity in EigenLayer’s restaking design. The same pattern applies here: when governance tokens control financial outcomes, the governance process becomes an attack surface. The UMA voting mechanism is susceptible to flash loans and vote buying. The banks’ restriction does not address this. It only shuts one door while leaving the oracle window open.
Kalshi, being regulated, relies on a centralized dispute process. Less elegant, but less vulnerable to economic manipulation. Security is the absence of assumptions. Kalshi assumes the regulator will enforce truth. Polymarket assumes the token holder will act honestly. One assumption is backed by law. The other is backed by game theory. In a bear market or low-stakes event, game theory often fails.
Contrarian: What the Bulls Got Right
The bulls would argue that this restriction is a bullish signal. It validates prediction markets as legitimate financial instruments. Banks only ban things that matter. The move also acknowledges that information from prediction markets is valuable—otherwise, insider trading would not be a concern. Moreover, the restriction only applies to employees. The broader retail user base remains unaffected. Polymarket’s daily volume during the 2024 US election cycle reached $300 million. Post-election, it has declined to $20–30 million. The loss of institutional traders might even improve market efficiency by removing noise from traders with skewed incentives.
There is also a network effect argument. If Polymarket becomes the trading venue for those without compliance constraints, it may attract more genuine information traders—the long tail of individuals with special knowledge but no corporate affiliation. That could make predictions more accurate, not less.
But this bull narrative ignores a critical variable: regulatory path dependency. Once the largest banks define a platform as “risky,” the SEC and CFTC take notes. The insider trading playbook is well-established: first internal bans, then industry guidance, then enforcement actions. We saw this with ICOs in 2018, with DeFi lending in 2021, with staking products in 2023. The pattern is predictable. The banks are not acting in isolation. They are signaling to regulators: we have identified a gap. Now close it.
Takeaway: The New Perimeter
The geometry of prediction markets is being redrawn. Polymarket’s permissionless model will survive only if it can prove that insider activity is negligible or can be detected through on-chain analysis alone. That is a hard proof to generate without KYC. Kalshi, already inside the regulatory perimeter, will likely absorb any institutional flow. The trade-off is clear: privacy for compliance, but compliance for longevity.
Zero trust is not a policy; it is a geometry. The new perimeter is not a technology stack. It is a jurisdictional boundary. Prediction markets will continue to exist, but the line between “free market” and “regulated exchange” will become sharper. For traders, the question is no longer “which platform has the best liquidity?” but “which platform will still be legal in six months?”
