MIT Study: AI Weakens Fake-News Detection as Kalshi Cracks Down on Insider Trading

Crypto News

A four-week MIT Media Lab study has raised fresh concerns about how heavily traders and researchers lean on AI to verify information across the blockchain ecosystem and beyond. Tracking 67 participants across 7,203 AI conversations and 4,536 news-authenticity judgments, researchers found that AI assistance lifted misinformation-detection accuracy by 21% in the moment. Yet when the same users later evaluated fresh headlines and images alone, their unassisted performance fell by 15.3 percentage points. The system paired GPT-4o with Google Search, while Claude 3.5 Sonnet analyzed the dialogue, suggesting reliance on chatbots may shift beliefs without building durable critical-thinking skills.

In trading infrastructure, Shotgun.fun launched as a high-performance terminal returning up to 100% of trading fees to users, directly challenging a fee model that has quietly extracted billions across decentralized exchange activity. Cashback begins at 50% and scales with volume, while the platform stays fully non-custodial through Turnkey key encryption. The product ships with real-time launch tracking, copy-trading discovery, one-click execution, limit orders and multi-wallet management. It is led by Miguel Loures and Pedro Maurício, the founders behind Pulsar Finance, a portfolio manager backed by Delphi Ventures that surpassed one million users before its acquisition by Terraform Labs.

Kalshi tightened its surveillance regime, announcing it will now require some users to disclose their employers before trading in markets flagged for elevated insider or manipulation risk. The federally regulated exchange said the policy took effect immediately, following recommendations from an independent Surveillance Audit Committee that reviewed its enforcement systems and trading controls. Kalshi reported blocking more than 100 potential insider trades in the first quarter using new screening tools, opening over 150 investigations, referring more than 20 cases to law enforcement and issuing five disciplinary actions. A new risk-scoring system now grades markets on insider-trading exposure, importance, regulatory concern and national-security implications.

The MIT findings carry weight precisely because the decline was driven by a reduced ability to spot fabricated content rather than genuine material. Because the experiment used the older GPT-4o and Claude 3.5 Sonnet, researchers cautioned that newer systems such as GPT-5.5 or Claude Opus 4.8, with stronger reasoning, might produce different outcomes. The timing is notable: social platforms including X are simultaneously working to curb AI-generated war footage, underscoring how synthetic media and verification tooling are advancing in tandem. For a market already prone to coordinated narratives, the prospect of outsourced judgment weakening independent scrutiny is a structural risk worth watching.

Kalshi's overhaul lands amid intensifying scrutiny of prediction markets broadly. A Yale and London Business School paper analyzing Polymarket activity from 2023 to 2025 found that just 3% of traders accounted for most price moves, a concentration that invites manipulation in thinly traded venues. The study spotlighted a U.S. Army Green Beret arrested in April over $400,000 in bets tied to a Venezuela raid he participated in, followed a month later by the arrest of a Google engineer for alleged insider trading. These cases illustrate why regulated venues are racing to layer identity verification and employment screening onto event-based wagering before reputational damage spreads.

Shotgun's pitch also targets the predatory dynamics that have drained retail participants, noting that insider wallets have extracted hundreds of millions across recent token launches. By surfacing those wallets and enabling users to view and mirror their moves in real time, the terminal frames transparency as a competitive edge in a crowded field of execution tools. A referral program offering up to 50% revenue share across five layers deepens the incentive to migrate volume. As DeFi trading consolidates around aggregators and automated market maker routing, fee rebates and insider visibility are emerging as the new battleground for trader loyalty.

Read together, these developments trace a single arc: the contest over trust and integrity as automation reshapes how participants judge information and execute trades. COINOTAG's aggregate market data underscores the fragility of that backdrop, with the Fear and Greed Index pinned at 9 out of 100 in Extreme Fear territory and total crypto market capitalization near $1.76 trillion. Bitcoin dominance sitting at 70.4% signals capital huddling into perceived safety while risk appetite for speculative venues thins. In a tape this defensive, the premium on verifiable data, transparent fee structures and credible surveillance is rising precisely when synthetic content and concentrated insider flows threaten to erode it fastest.