United States v. Ding tests the limits of economic espionage

Where prosecutions for the federal offense of economic espionage are concerned, timing and sometimes an accused's ineptitude can be everything, as illustrated by a recent decision in United States v. Ding, N.D. Cal. Case No. 24-cr-VC.

For background: In February 2023, the Departments of Justice and Commerce launched the Disruptive Technology Strike Force to combat the illegal exporting of sensitive technology to nation-state adversaries. In May 2023, the government announced the task force's first fruits, which included two criminal cases charging former software engineers with stealing software and hardware source code from U.S. tech companies to market it to Chinese competitors. The creation of this multi-agency federal task force and its early criminal charges may be a harbinger of more such cases.

The task force's cases featured charges under the Economic Espionage Act of 1996 (EEA), which criminalizes two related types of conduct: theft of trade secrets (18 U.S.C. § 1832) and economic espionage (18 U.S.C. § 1831). The critical difference between the two offenses is the required intent. Unlike the former, the latter requires proof that the defendant intended or knew the theft would benefit a foreign government, instrumentality or agent--which often raises interesting questions of proof, as demonstrated in the more recent Ding case.

Linwei Ding was a mid-level Google software engineer who snapped more than 2,000 screenshots of confidential Google documents containing AI trade secrets and pasted them into Apple Notes to evade detection. Ding then tried to launch his own AI startup in China. Ding was arrested in January 2024 and charged in the Northern District of California with seven counts of theft of trade secrets and seven parallel counts of economic espionage. A jury convicted Ding on all counts earlier this year. After multiple rounds of post-trial briefing, however, the district court (Judge Vince Chhabria) on Aug. 20 affirmed the theft-of-trade-secrets verdicts but acquitted Ding on every count of economic espionage.

Chhabria recognized that to prove theft of trade secrets, the government must show only that Ding intended to benefit himself, and he found there was ample evidence of that. But to prove economic espionage, he noted, the government must do more than that. It must also show that Ding intended or knew his conduct would benefit the Chinese government itself--not merely that it would benefit China's economy "in a way that might generally flow from engaging in economic activity there." As his order succinctly put it, "Intent to start a private company in China is not enough."

According to Chaabria, the government's economic espionage case against Ding foundered on an issue of timing. Ding secretly uploaded the Google documents in four batches between May 2022 and April 2023. But there was no evidence Ding was in contact with anyone from the Chinese government until September 2023, when he interacted with several Chinese-controlled entities to promote his AI startup. For Chhabria, the key factual question for the economic espionage charges was whether a jury could find beyond a reasonable doubt that Ding had intended to benefit those government entities at the time he stole the documents. He allowed that later conduct can sometimes be used to prove intent when an offense is completed: steal the documents on Monday and hand them to Chinese officials on Friday, and "nobody would question" the inference. But "at some point . . . it becomes too speculative," and for Chhabria, five months was too long. 

As an aside, one may well wonder how this "temporal disconnect between the uploads and Ding's engagement with the Chinese government," so critical to the court's ruling, escaped the litigants' notice at trial. Chhabria's order posits that this resulted from Ding's defense that he wasn't really trying to steal the documents at all. Since Ding was swinging for the fences on all counts, the order surmises, an alternative defense that Ding wasn't trying to steal to benefit the Chinese government--applicable only to the economic espionage counts--was more of a bunt by comparison, and therefore not in the defense game plan.

In briefing on Ding's Rule 29 motion, the government unearthed from its many trial exhibits a promotional document from April 2023 indicating that potential customers for Ding's AI company included "national supercomputing centers," which, according to prosecutors, were Chinese government entities. The government also pointed to evidence in the trial record that Ding downloaded the stolen files from his Google cloud account to a personal laptop when he feared detection in December 2023, after he interacted with Chinese-controlled entities.

But this was not enough to persuade Chhabria. He acknowledged that the evidence could have supported the espionage convictions. A reasonable jury could conclude, for example, that Ding committed theft of trade secrets in April 2023 and later committed economic espionage in December 2023. But he declined to consider these theories of guilt because the government did not argue them to the jury. "It's black-letter law that the government can't save a conviction on appeal by using a theory it didn't present to the jury," wrote Chabbria.

This rule appears well-supported when the jury has not been instructed on the theory the government belatedly relies on. But if the jury has been properly instructed, there is at least some authority suggesting that a defendant's conviction may be upheld even though the prosecution's closing arguments do not emphasize the particular factual theory that ultimately supports the conviction on sufficiency-of-the-evidence review. See Musacchio v. United States, 577 U.S. 237, 244 (2016). On this point, the court's ruling may be vulnerable on appeal, should the government decide to pursue one.

A further aspect of Chabbria's ruling may interest practitioners in both criminal and civil economic espionage cases. That is the way Ding's own characteristically "ham-fisted" conduct helped save his skin. As just one example, Google apparently first started to suspect Ding of stealing its trade secrets after he asked an intern to swipe into the office with his security badge so that he could hide the fact that he was actually in China promoting his non-existent company. More generally, although Ding ultimately squirreled away nearly 2,000 documents filled with confidential AI information from Google, he seemed to have no idea what to do with the information in them. As Chabbria's order describes it, "[E]veryone involved at trial seemed to take the view that Ding lacked the talent, skill, knowledge, or business acumen to create an AI supercomputing company." For Chhabria, the bungling nature of the way Ding committed theft of trade secrets suggested Ding "lack[ed] a real plan for his product and company," thus undercutting the notion that he stole Google's trade secrets with the intent or knowledge that his offense would benefit the Chinese government. While a person's inability to achieve a Quixotic plan does not always suggest a lack of intention to carry it out, Ding's case teaches that, when defending an economic espionage case, incompetence may sometimes be converted into an asset.