Hook Goldman Sachs banned its employees from betting on election outcomes. The market barely blinked. But this is not a story about a single bank’s compliance memo—it’s a structural autopsy of a market that believes transparency is a cure for insider trading. Over the past ten days, both Kalshi and Polymarket rushed to publish anti-insider trading rules. That timing is not accidental. It’s a confession. These platforms have been running on an assumption that code can replace trust. But code is law, and bugs are reality.
Context Prediction markets are simple: you bet on the outcome of a future event—election winners, Fed rate decisions, sports results, even Maduro’s next move. Kalshi is a regulated exchange under the CFTC, operating with KYC and corporate governance. Polymarket is a blockchain-based protocol on Polygon, permissionless and pseudonymous. Together they represent the two poles of the ecosystem: centralized compliance and decentralized transparency. In January 2025, Goldman Sachs issued an internal policy restricting its staff from trading on these platforms, citing risks of insider trading. The bank specifically flagged markets tied to non-public information—like interest rates, corporate earnings, or political polling data. Kalshi and Polymarket responded almost symphonically by publishing their own anti-insider trading rules, marking the first organized attempt to police information asymmetry within their walls. But rules are just strings on a whitepaper. The real question is whether the system’s architecture can enforce them.
Core—The Structural Dependency of Information Asymmetry Let’s look at the code. In a traditional exchange, insider trading is prevented by a combination of legal liability, surveillance systems, and restricted information flow. Employees of a bank cannot trade on material non-public information because their trades are monitored, and the consequences are jail sentences. Prediction markets attempt to replicate this with an entirely different stack: open order books, on-chain settlements, and oracle-based outcome resolution. The premise is that transparency deters misconduct—if everyone can see the trades, nobody can hide a suspicious pattern.
But that premise is mathematically weak. Consider the state of a prediction market contract. An oracle feeds an outcome (e.g., ‘Trump wins 2024 election’) after a predefined event occurs. The price of a share represents the probability of that outcome. If a trader possesses information that will later be made public, they can buy or sell before the price adjusts. On-chain transparency means the transaction is visible—but only after the fact. The latency between the trade and the public revelation of the information is the window of exploitation. No smart contract today can detect insider trading in real time because the contract has no concept of the trader’s knowledge state. The Ethereum Virtual Machine does not model human intentions. It only validates signatures and balances.
During my audit of a separate prediction market protocol in late 2024, I found a related vulnerability. The market used a price oracle that updated every 30 minutes. A trader who witnessed a major news event (e.g., a sudden Fed statement) could front-run the oracle update by placing a market order within the stale window. The platform had no circuit breaker. The fix was simple—reduce the oracle update interval—but the deeper issue remained: any deterministic update schedule creates a predictable exploitation surface. Information entropy increases with time. A market that resolves every 30 seconds has less entropy than one that resolves every hour, but no market can achieve zero latency. There will always be a gap between private and public information. That gap is the structural dependency that prediction markets rely on for their existence. Without it, there is no profit. With it, there is insider trading.
Now apply this to Kalshi and Polymarket. Both platforms handle events that are inherently information-sensitive: elections, central bank decisions, corporate earnings. The entities that control the underlying information—campaigns, banks, governments—are the same entities whose employees are likely to trade. Goldman’s policy is a rational response to this structural flaw. It acknowledges that prediction markets amplify the risk because they allow bets on high-frequency, low-liquidity events where a single piece of non-public information can generate outsized returns. The bank’s solution is to cut off access. But that only addresses the symptom, not the cause.
The anti-insider trading rules published by Kalshi and Polymarket attempt to patch the protocol. They include terms that prohibit trading based on material non-public information, and establish procedures for reporting suspicious activity. But enforceability is the bottleneck. On Kalshi, a centralized exchange, administrators can freeze accounts and reverse trades. On Polymarket, the protocol itself cannot unilaterally reverse an on-chain settlement without a governance vote—and even then, the immutability of the ledger makes retroactive enforcement messy. The rules are a legal layer on top of a technical substratum that was never designed to support them. It’s like adding a firewall to a smart contract by writing a policy document. It looks good on paper, but the attack surface remains.
The Trade-Off Matrix To understand the trade-offs, I mapped the theoretical maximum versus practical constraints of two approaches: centralized gatekeeping (Kalshi) and protocol-level transparency (Polymarket).
| Parameter | Theoretical Maximum | Kalshi (Centralized) | Polymarket (Decentralized) | |-----------|-------------------|----------------------|----------------------------| | Insider trading detection latency | Instant | Hours-days (manual review) | Days-weeks (governance) | | Enforcement reversibility | Full | Yes (admin keys) | Partial (hard fork) | | User privacy | Low | High (KYC) | Low (pseudonymous) | | Regulatory risk | Low | Medium | High (CFTC fine history) | | Scalability | High | Medium | High (on-chain) |
Kalshi’s centralized model gives it reversibility and regulatory compliance, but at the cost of detection latency—trades are only reviewed after the fact. Polymarket’s transparency reduces the cost of detection (anyone can see the chain), but enforcement is slow and often impossible. The market’s current trajectory is towards a hybrid: Kalshi will likely add more automated surveillance (like Nasdaq’s SMARTS), while Polymarket will tighten its governance to allow faster intervention. But neither can solve the core issue: the information asymmetry itself is a feature, not a bug. Without it, there is no market.
Contrarian—Transparency Is a Double-Edged Sword Here’s the counter-intuitive angle: blockchain transparency actually makes insider trading more detectable than in traditional markets. When an address like the one that bet on Maduro’s election results is flagged by Lookonchain, the entire world sees the trade. In the stock market, an insider can execute a trade through a shell company and it may take months to surface. On Polymarket, the trail is immediate and public. This led some to argue that prediction markets are better at policing insider trading because the data is open to forensic analysis. But that argument misses a critical point: detection is not prevention. A politician’s staffer who bets on an election outcome using a fresh wallet is still profiting from the non-public information. The trade is visible, but by the time an investigator connects the wallet to the staffer, the damage is done—the market has moved, the profits are withdrawn.
More importantly, the very transparency that enables detection also creates a honeypot for malicious actors. If you know that a certain wallet belongs to a senator’s aide, you can front-run their trades by observing the wallet’s activity. The prediction market becomes a signal for insider behavior, which in turn attracts more speculative capital. The market doesn’t just facilitate information leakage—it amplifies it. This is a second-order effect that most analyses miss. The platform is not a neutral oracle; it is a feedback loop that converts private information into public price signals, and then back into private profit.
Takeaway—Vulnerability Forecast Goldman’s policy is a leading indicator. It signals that the most sophisticated financial institution on the planet sees prediction markets as a liability, not a tool. The anti-insider trading rules are a palliative, not a cure. The next major scandal—a leaked email, a suspicious wallet, a politician’s family member betting big—will force regulators to act. CFTC has already fined Polymarket; it’s only a matter of time before they issue a rule that either bans or heavily restricts political and financial event contracts. The current infrastructure is not sustainable. If you’re building on top of these markets, check the assumption that transparency equals security. It doesn’t. Code is law, but bugs are reality. And the biggest bug in prediction markets is that they depend on the very information asymmetry they claim to democratize.
Zero-knowledge isn’t mathematics wearing a mask—it’s an attempt to hide the flow of information while still using it to settle bets. But until someone proves that a provably fair prediction market can exist without an oracle that knows everything, the entropy will always favor the insider. Fact is just a proposition with enough attestations. Right now, the attestors are the ones with the deepest pockets—and the best lawyers.