US weighs a 'Find My' for AI chips: pings, certificates, fingerprints

The US restricted exports of certain advanced chips to China in 2022, yet IEEE Spectrum notes those chips have kept turning up in China throughout the AI boom. On 1 October 2026 the US arrested a man on charges of sending over $300 million worth of GPU-equipped servers to China, using paperwork meant for shipments to Malaysia and Singapore. The article's point: the more the US tries to limit the movement of its AI chips, the harder they are to locate precisely.

The response under discussion is location verification. Last March, the US House Foreign Affairs Committee passed a bill requiring it for advanced AI chips, and ahead of the US-China summit in late September legislators called for a floor vote on that and related bills. If a requirement passed, chipmakers could comply by building tamper-proof root of trust hardware, implementing cryptographic signing and transmission, or possibly fingerprinting individual chips. The idea resembles the 'Find My' feature that locates lost Apple devices.

The lead candidate is ping-based location verification (PLV). It measures how long a ping takes to travel from the chip to several trusted 'landmark' servers around the world and back, and deduces the chip's location from that. A September report from the Center for Security and Emerging Technology (CSET) says a combination of approaches, such as PLV supplemented by physical inspections, would be feasible and robust. Kyle Miller, a senior research analyst at CSET and coauthor of the report, says manufacturers could build PLV into the hardware of new chips, while a firmware update could reach chips already in circulation. Not every chip needs tracking: Jacob Feldgoise, a senior data research analyst at CSET and coauthor, says the priority should be the highest-end data center GPUs.

A related scheme uses encrypted 'sovereignty certificates'. Lucid Computing, a secure computing firm, has published a draft specification describing them as cryptographic credentials that give verifiable, unforgeable proof of a computing device's physical location at a specific point in time. Kristian Rönn, Lucid's CEO and cofounder, explains that a chip holding a private key can sign messages, a landmark server receives and times them, and the result is 'a radius of where the chip has to be'. Many modern chips have trusted execution environments that support a hardware root of trust for this. Rönn sees the certificates as essential tools in future and likens them to a 'lie detector' for data centers.

Miller says PLV could deter some smuggling, because smugglers know a powered-on chip starts pinging landmark servers; when a chip is flagged in an unintended place, PLV data also gives a better sense of when and where it moved. But PLV only reveals diversion after it has happened. It cannot track offline chips in storage, in transit or not yet activated. If a chip does not answer a ping, the system cannot tell whether the cause is innocent or nefarious. That is why Miller argues for physical audits of data centers: PLV 'generates leads, and you track down those leads'.

The method can also be attacked. An adversary with physical access to a chip could stop it sending signals, or extract its cryptographic keys and spoof a signal from the intended location. Some security researchers are therefore exploring hardware fingerprints instead of keys; one study demonstrated a proof-of-concept function that produced unique fingerprints for Nvidia H200 GPUs from manufacturing variations and random numbers as seeds. Rönn says other fingerprinting should be supported because not all chips have a hardware root of trust, and that tamper-proofing and defense in depth matter too.

The CSET report warns that PLV is costly. If China gains the ability to make its own advanced AI chips, Miller says, tracking US chip exports becomes less of a strategic priority. Hanna Dohmen, a senior research analyst at CSET and coauthor, points to another gap: Chinese companies can legally access AI chips abroad remotely through cloud providers, and such chips may sit in data centers in Southeast Asian countries like Malaysia that increasingly serve Chinese firms. Whether or not a 'Find My' for chips becomes law, Feldgoise expects tighter conditions on advanced chip exports, because compute is increasingly seen as a strategic resource.

Key facts

On 1 October 2026 the US arrested a man charged with sending over $300 million worth of GPU-equipped servers to China using paperwork meant for Malaysia and Singapore.
Last March the House Foreign Affairs Committee passed a bill requiring location verification for advanced AI chips; legislators asked for a floor vote ahead of the late-September US-China summit.
Ping-based location verification times pings between a chip and trusted landmark servers; a September CSET report says it works best supplemented by physical inspections.
PLV only uncovers diversion after the fact, cannot track offline chips, and is costly per the CSET report; Miller says it 'could deter some smuggling'.
Alternatives include Lucid Computing's draft sovereignty certificates and hardware fingerprinting, shown so far as a proof of concept on Nvidia H200 GPUs.

Why it matters

Export controls on advanced chips have been in place since 2022, but the article says chips keep reaching China anyway, and the October arrest over more than $300 million of servers is the latest example. Location verification would give the US a way to see where controlled chips actually are, and a House committee has already passed a bill that would require it. The article also frames compute as a strategic resource, and Feldgoise expects due diligence requirements on chips to tighten over time.

Who it affects

Chipmakers would have to build in the capability: tamper-proof root of trust hardware, signing and transmission, or possibly fingerprinting. Owners of chips already in circulation could receive it by firmware update, per Miller. Data center operators, including those in Southeast Asia such as Malaysia, could face physical audits. US lawmakers and export-control enforcers would get leads on diverted chips. Chinese firms, which can legally use chips abroad through cloud providers, are affected in a different way, since that route sits outside what location tracking addresses.

How to use it

There is nothing to install or buy here. This is a policy and engineering debate. For those building or regulating, the article lays out the pieces: PLV via landmark servers, signed messages from a key held on a chip with a hardware root of trust, and fingerprinting for chips without one. CSET's authors advise pairing PLV with physical inspections, and Lucid Computing's draft specification for sovereignty certificates has been published.

How solid is it

The article is by Rina Diane Caballar, an IEEE Spectrum contributing editor, and rests on a September CSET report plus interviews with three CSET analysts (Miller, Feldgoise, Dohmen) and Lucid Computing's CEO, Rönn. The arrest and the committee vote are stated as facts. The proposals are not law: the bill only passed committee, and the article describes a floor vote as something legislators called for. The fingerprinting study is described only as a proof of concept, and the closing editor's note says the sources on GPU fingerprinting are still thin. Rönn leads the company that published the sovereignty certificate draft.

Risks and caveats

PLV can only find diversion after it has happened and cannot track chips that are offline, in storage, in transit or not yet activated. A chip that does not answer a ping leaves the system unable to tell an innocent cause from a nefarious one. An adversary with physical access could block signals, or extract keys and spoof a location. The CSET report warns PLV is costly. If China builds its own advanced chips, Miller says tracking US exports matters less. Remote cloud access to chips abroad is currently legal, per Dohmen. Miller says PLV could deter some smuggling, which is short of stopping it.

“Currently, Chinese companies are legally able to remotely access AI chips abroad through cloud service providers”

— Hanna Dohmen, senior research analyst at CSET and coauthor of the report