Tao Li served as a program officer at the National Science Foundation from 2015 to 2017, with authority over grant awards in hardware, software and algorithms. During that period he participated in two Chinese talent plans, according to a report by the NSF’s Office of Inspector General.

While in that role, Li was involved in awarding US research grants to at least seven other scientists who allegedly also failed to disclose their participation in one of those plans — a program called Dragon Star.

The FBI raised concerns about Li’s suspected Dragon Star involvement with NSF leadership in 2019, according to a former CIA analyst who was in the meeting. The NSF debarred Li from federal funding for six years in May 2026. The Justice Department declined a criminal referral in 2024.

By the time NSF acted, Li was running an artificial intelligence research center at a military-affiliated university in central China.

The Timeline Is the Story

Strip out everything else and read the dates:

  • 2015–2017 — Li holds NSF program officer authority while allegedly in two Chinese talent plans
  • 2019 — FBI briefs NSF leadership on suspected Dragon Star involvement
  • 2024 — DOJ declines to prosecute
  • May 2026 — NSF imposes a six-year funding debarment
  • August 2026 — reporting surfaces that he already leads an AI center at a PLA-affiliated institution

That is seven years from federal law-enforcement notification to administrative sanction. The sanction is exclusion from a funding stream the subject no longer needs, imposed after he relocated to the beneficiary country and took a position in the exact research domain he had oversight of.

Investigators also found that Li deleted work-related files from Dropbox, including a folder whose title referenced the National Science Foundation. Deletion of records under investigation is normally the fact that accelerates a case. Here it sat inside a process that took seven years.

Program Officer Is the Position That Matters

Most research-security enforcement targets grant recipients — a professor who took Thousand Talents money and did not put it on the disclosure form. That is a compliance failure with a bounded blast radius: one lab, one portfolio of work.

A program officer is a different position entirely. The role carries visibility into the full submission pipeline — proposals that were never funded, reviewer comments, agency priorities, which lines of research the United States is choosing to invest in and which it is passing over — and influence over which proposals advance.

The unfunded-proposal pipeline is the underrated asset. A rejected NSF proposal is a detailed technical roadmap, written by a competent researcher, describing work that will likely proceed anyway under other funding. It is never published. Portfolio-level visibility across that pipeline is a strategic view of American research direction that no amount of open-source collection reproduces.

The seven other scientists Li allegedly helped fund, who allegedly also concealed Dragon Star participation, compound this. If accurate, that is not one compromised individual — it is a funding decision-maker allocating public research dollars toward a network with an undisclosed shared affiliation.

Dragon Star and the Disclosure Regime

The FBI’s standing caution about Chinese talent plans is that they “can incentivise members to steal or misappropriate foreign knowledge or technology.” Dragon Star is less prominent than Thousand Talents but functions in the same category — the NSF’s own rules now treat participation in a Malign Foreign Talent Recruitment Program as disqualifying for federal funding, and require affirmative certification from covered personnel.

Those requirements postdate Li’s tenure. The rule change is the system responding to exactly this class of case — which is worth stating plainly, because it is the part that worked. What did not work is the enforcement clock.

What Seven Years Buys

The value proposition of research-security enforcement is deterrence, and deterrence is a function of speed and consequence. This case supplies a weak version of both: a funding debarment against someone with a foreign academic appointment, seven years after the FBI’s warning, with no criminal exposure.

The DOJ declination in 2024 is the pivot point and the least explained element in the public record. Without the underlying reasoning, the reasonable inferences range from evidentiary insufficiency to the collapse of the China Initiative and its aftermath, and none of them can be asserted from what has been published.

What can be asserted is the outcome. A former federal grant-making official is now directing AI research at a military-affiliated Chinese university, and the sanction the United States imposed is that he may not apply for American grants until 2032.

Sources