AI in quant trading

Quant trading desks are starting to use artificial intelligence for live market calls, not just research and backtesting. (congress.gov) Quant trading is automated investing: firms turn data into rules, then let software place trades in stocks, futures, options, or currencies in milliseconds. A July 2025 Congressional Research Service brief said artificial intelligence can now help identify, execute, and back-test trading strategies, while newer generative systems can process large volumes of unstructured data. (congress.gov) That matters because markets produce two kinds of information at once: structured price data and messy text, audio, and images. A July 2024 survey from researchers at Columbia University and New York University said large language model trading agents are being built to read those mixed inputs quickly enough to support trading decisions. (arxiv.org) In practice, that means a model can scan earnings calls, shipping updates, policy statements, satellite images, or social posts, turn them into signals, and feed those signals into an execution system. The Congressional Research Service said generative artificial intelligence has enabled investment firms to use unstructured information to sharpen analytic trading tools. (congress.gov) The public evidence is still uneven. Researchers have published surveys and backtests, but the strongest claims about live profits, win rates, or specific commodity trades often come from traders, vendors, and social-media posts rather than audited fund disclosures. (arxiv.org) Regulators have moved in alongside the marketing. On December 5, 2024, the Commodity Futures Trading Commission issued a staff advisory telling registered firms that existing rules still apply when they use artificial intelligence in CFTC-regulated markets. (cftc.gov) The Securities and Exchange Commission has also focused on how firms describe these systems. On March 18, 2024, the SEC charged Delphia and Global Predictions with making false and misleading statements about their use of artificial intelligence, and the firms agreed to pay $400,000 in civil penalties. (sec.gov) The SEC then held a public roundtable on March 27, 2025 on the risks, benefits, and governance of artificial intelligence in finance. That session put trading firms, regulators, and compliance teams in the same conversation as firms test whether language models belong inside investment workflows. (sec.gov) The bigger change is less about one chatbot picking trades and more about the plumbing around it. The newest systems are being used as research assistants, code generators, data cleaners, and signal-ranking tools that sit between raw information and the order sent to market. (congress.gov; arxiv.org) For now, the industry is in a familiar phase: faster adoption than disclosure. Firms are testing how much decision-making to hand to models, while regulators are pressing them to prove what the systems actually do. (cftc.gov; sec.gov)