Prosecuting Insider Trading In The AI Era

A version of this article originally appeared in Business Law Today, the ABA Business Law Section's digital magazine offering in-depth articles, quick summaries of the month's key developments by practice area, checklists and other practical tools, and more.

In Brief

• To prove insider trading, the government must establish that a defendant traded securities "on the basis" of material nonpublic information ("MNPI") in violation of a duty of trust. Artificial intelligence could make it easier to create a defense to a charge of insider trading.
• A person with MNPI about a particular public company could query AI about the company's stock, obtain an analysis that supports buying the stock, and then argue that they traded based on the AI analysis, not "on the basis" of MNPI.
• Some courts follow the "knowing possession" standard, requiring proof that a defendant traded while knowingly possessing MNPI. Other courts follow the "use" standard, requiring proof that the defendant traded because of the nonpublic information. The Supreme Court has not ruled.
• Under the "mosaic theory," a trader who gathers information from multiple public and nonpublic sources, analyzes the data, and makes an independent decision cannot be liable for insider trading if no one piece of information qualifies as MNPI.
• Successfully using AI-assisted trading as a defense to an insider trading prosecution will depend on how a court interprets the phrase "on the basis of." A "use" standard will be helpful to an investor; a "knowing possession" standard will not.
• Either way, AI could provide the ready-made defense that may be enough to create reasonable doubt.

To prove insider trading, the government must establish that a defendant traded securities "on the basis" of material nonpublic information ("MNPI") in violation of a duty of trust. In many insider trading investigations, those under the microscope will try to show the government that they traded for a reason other than on the basis of MNPI. The success of such a defense depends in part on the interpretation of the language "on the basis of." Since the 1990s, there has been a circuit split, still unresolved by the U.S. Supreme Court, on how to interpret these four words. Some courts follow the "knowing possession" standard, under which the government need only prove that a defendant traded while knowingly possessing MNPI. Other courts follow the "use" standard, requiring the government to prove that the defendant traded because of the nonpublic information.

Like everything else, artificial intelligence ("AI") may soon change the way those suspected of insider trading defend themselves. AI, which investors and analysts increasingly rely on to trade stocks and analyze data, could make it easier for an individual investor to create a defense to insider trading. Imagine that an individual gets inside information about a publicly traded company. However, before trading, that person queries AI about the stock and obtains a comprehensive bespoke analysis supporting buying the company's stock. AI makes it possible, if not easy, for someone possessing MNPI to create a ready-made defense should the government pursue charges for insider trading—and it may be enough to create reasonable doubt.

The Rise in AI-Assisted Trading

Investors and analysts are increasingly using AI in stock trading to analyze data, generate stock picks, and automate trading decisions. And for obvious reasons: AI offers advanced algorithms that can process large volumes of real-time data in a fraction of the time it would take even a seasoned analyst to do the same. AI can also analyze complex patterns in stock prices, financial statements, economic indicators, and even news articles to make predictions with higher accuracy than traditional methods—and all without human intervention.

New trading firms like XTX Markets and Tiger Brokers purportedly use AI to execute millions of trades daily, and there are AI-driven funds like Pictet that leverage AI to try to improve returns. Beyond the professionals, AI trading tools are also becoming available to retail investors. Experts are predicting that this could create a "seismic transformation" in the entire stock market.¹

Overview of Insider Trading Law

Insider trading is commonly understood as buying or selling securities on the basis of MNPI in breach of a duty. Importantly, there is no federal statute that explicitly bans insider trading. Rather, the law of insider trading has evolved through common-law concepts of fraud and deceit.

The "classical theory" of insider trading refers to a case of a corporate insider who owes a fiduciary duty to the company or its shareholders and breaches that duty by trading on the basis of MNPI.² The "misappropriation theory" expands liability by assigning fiduciary duties to certain corporate outsiders.³ United States v. O'Hagan held that although the defendant was not a corporate insider of the acquirer, he owed a duty to his firm's client, the source of the information, not to trade on MNPI.⁴ Liability for insider trading can attach both to a person that trades on MNPI and an individual that provides a tip in breach of a duty for the personal benefit of the tipper.⁵

The Debate over "Use" Versus "Knowing Possession"

Under both the classical and misappropriation theories of insider trading, a defendant must purchase or sell securities in reliance on, or "on the basis of," MNPI.⁶ Courts have struggled to unify behind a common interpretation of what "on the basis of" means, and two standards have emerged: the "use" standard and the "knowing possession" standard. In other words, does the government have to prove that an insider actually used MNPI to trade, or is it sufficient that the defendant knowingly possessed MNPI at the time of the trade?

A series of appellate decisions in the 1990s created a circuit split on this question. First, in United States v. Teicher, the defendant appealed his insider trading conviction, arguing that he had traded on the basis of public information that led him to become interested in the stocks at issue rather than the inside tips he had received.⁷ Teicher argued that the jury instructions wrongly permitted a conviction "[b]ased upon the mere possession of fraudulently obtained material nonpublic information without regard to whether this information was the actual cause of the sale or purchase of securities."⁸ The U.S. Court of Appeals for the Second Circuit rejected the argument and held that the government solely need to prove "knowing possession" of MNPI to obtain a conviction for insider trading.

Six years later, the U.S. Court of Appeals for the Eleventh Circuit reached the opposite conclusion in SEC v. Adler.⁹ In that case, the court there held that to prove a defendant engaged in insider trading, the U.S. Securities and Exchange Commission ("SEC") needed to show that the defendant "used" the nonpublic information to trade because a broader "knowing possession" standard would capture cases that did not actually involve fraud. To address the concern that a "use" standard would impede the government's prosecution of insider trading, the court created a "burden-shifting" framework:

Shortly after Adler, the U.S. Court of Appeals for the Ninth Circuit applied the "use" standard in a criminal case. In United States v. Smith, the court found that "use" rather than "knowing possession" was more consistent with the scienter requirement of § 10(b) of the Securities and Exchange Act of 1934 and Rule 10b-5.¹¹ The Ninth Circuit also declined to apply the "strong inference" that the Adler court introduced because criminal cases prohibit presumptions of fact.¹²

SEC Rule 10b5-1

The SEC attempted to resolve this debate through rulemaking and proposed Rule 10b5-1 in December 1999, which was enacted in August 2000. It effectively adopted the "knowing possession" standard, providing as follows: "[A] purchase or sale of a security of an issuer is on the basis of material nonpublic information [about that security or issuer] if the person making the purchase or sale was aware of the material nonpublic information when the person made the purchase or sale."¹³

Despite the SEC's rulemaking, courts have inconsistently adopted the Rule 10b5-1 standard. For example, the Eleventh Circuit in Fried v. Stiefel Laboratories, Inc. ignored Rule 10b5-1 and cited Adler to conclude that a finding of liability required proof that the defendant had actually used MNPI to trade.¹⁴ The U.S. Court of Appeals for the Eighth Circuit similarly concluded that the government must prove that the defendant "actually used the information," without citing Rule 10b5-1.¹⁵ Lower courts have also differed on the deference afforded to the SEC's definition.

Critically, the Supreme Court has not weighed in, and there has not been much development in this law since the late 1990s.

AI-Assisted Trading as a Potential Defense in Insider Trading Prosecutions

A key defense to insider trading is that the defendant did not trade "on the basis of" MNPI. The defendant might claim that the trading decision was based on a factor independent of MNPI, and AI tools could provide that independent factor. For example, the defendant could establish that the trades were part of a routine investment strategy or diversification effort and were motivated by publicly available information aggregated by AI.

Relatedly, the use of AI could be part of a mosaic defense to insider trading. The "mosaic theory"¹⁶ is the view that a trader who gathers information from multiple public and nonpublic sources, analyzes the data, and makes an independent decision cannot be liable for insider trading if no one piece of information qualifies as MNPI. The defendant could argue that even if he had tidbits of nonpublic information, he used AI to supplement his knowledge and create a "mosaic" of information, and the entire picture provided the reason for his trades.

An investor who gets inside information about a stock could query AI about the stock before trading and then point to AI's analysis as the reason for its trades. Of course, ten years ago a defendant could accomplish the same goal by running Google searches for a stock before trading. But the sophistication of AI's capabilities and the rise in AI-assisted trading make it so much easier to have a cogent investment thesis.

The ability to successfully use AI-assisted trading as a defense to an insider trading prosecution will depend on how a court interprets "on the basis of." Under a "use" standard, an investor could show that it used AI, not the MNPI, as the basis for the trade. Under a "knowing possession" standard, AI-assisted trading would do little to support a defense if the government was able to show knowing possession of MNPI.

A key development in this area might be the Supreme Court's rejection of Chevron deference last year. In a post-Chevron world, courts may be more inclined to independently interpret § 10(b) without deference to Rule 10b5-1, and to conclude that the "use" standard is more consistent with the statute's proscription of deception and manipulation. Indeed, the government might have a harder time convincing a court that the "knowing possession" standard embraced by Rule 10b5-1 comports with Rule 10(b).

Conclusion

In sum, AI-assisted trading might present an obstacle to the government's prosecution of insider trading. At a minimum, assuming AI-assisted trading continues to expand, the government may have to consider how defendants could use this information when it assesses whether to bring an insider trading case.

Footnotes

1. Retail Stock Investors Can Now Imitate the Pros with AI Trading Tools, Bloomberg (June 10, 2025).

2. See Chiarella v. United States, 445 U.S. 222 (1980).

3. See United States v. O'Hagan, 521 U.S. 642 (1997).

4, Id.

5. See Dirks v. SEC, 463 U.S. 646 (1983).

6. O'Hagan, 521 U.S. at 651–52.

7. 987 F.2d 112 (2d Cir. 1993).

8. Id. at 119.

9. 137 F.3d 1325 (11th Cir. 1998).

10. Id. at 1337.

11. 155 F.3d 1051 (9th Cir. 1998).

12. Id. at 1069.

13. 17 C.F.R. § 240.10b5-1(b).

14. 814 F.3d 1288, 1295 (11th Cir. 2016).

15. United States v. Anderson, 533 F.3d 623, 631 (8th Cir. 2008).

16. In Dirks v. SEC, the Supreme Court appeared to recognize the legality of the mosaic theory of securities analysis, in which analysts obtain fragments of information from company insiders and then use those fragments to create a mosaic of information to value the company. 463 U.S. 646 (1983). "It is commonplace for analysts to ferret out and analyze information," the Court said, "and this often is done by meeting with and questioning corporate officers and others who are insiders." Id. at 658. However, since Dirks, a number of high-profile defendants, including Raj Rajaratnam, Rajat Gupta, and Doug Whitman, were unsuccessful in arguing the mosaic defense to insider trading.