on-chip governance mechanisms
In plain wordsFeatures built into AI chips that could check or enforce rules about how they are used — for example, proving which country a chip is in, or refusing to run without a valid license.
Hardware or firmware features in AI accelerators that allow a chip's use to be verified or constrained for policy purposes; also called hardware-enabled mechanisms (HEMs). Proposed functions include location verification (estimating where a chip is from the round-trip delay of signed messages to trusted servers), tamper-evident logging of workloads, remote attestation of the code and model being run (building on trusted execution environmentstrusted execution environment, n.: A sealed-off part of a computer chip that runs programs which the rest of the machine, even its operating system, cannot spy on or interfere with. and confidential computingconfidential computing, n.: Running computations inside a hardware-sealed compartment of a chip, so that data and AI models stay protected even from the cloud company or administrator operating the machine.), limits on how many chips can be networked into one cluster, and licensing schemes under which a chip needs periodically renewed authorization to keep operating at full speed. Advocates see them as a way to enforce export controls on AI chipsexport controls on AI chips, n.: Government rules that limit which countries and companies may buy the most advanced AI chips and the machines used to make them. The United States uses them mainly to slow China's progress in AI. after sale and to verify future international agreements on training; critics, including chipmakers, warn of new security vulnerabilities, perceived backdoors and lost customer trust, and engineers caution that robust tamper resistance is difficult. Some functions can be added through firmware or software on existing chips.
Location verification would let an exporter show that a licensed chip was still in the approved country.
Origin
Developed in research and think-tank work in 2023–2024, notably Yonadav Shavit's compute-monitoring proposal (2023), the Center for a New American Security report 'Secure, Governable Chips' (2024) and RAND's 'Hardware-Enabled Governance Mechanisms' (2024). The Chip Security Act, a bill introduced in Congress in 2025 and still debated in 2026, and America's AI Action PlanAmerica's AI Action Plan, prop. n.: The Trump administration's 2025 strategy for winning a global AI race: cutting regulation, building data centers and power plants quickly, and exporting American AI technology to allies. (2025) took up location verification for exported chips.
Usage note
Commercially sensitive: in 2025, after Chinese regulators questioned whether its H20 chips contained tracking features, Nvidia publicly denied that its chips had kill switches or backdoors.
Related
- compute governancecompute governance, n.: Using control over the computer chips and data centers that AI depends on — who can buy them, where they go, how they are used — as a way to oversee powerful AI.
- export controls on AI chipsexport controls on AI chips, n.: Government rules that limit which countries and companies may buy the most advanced AI chips and the machines used to make them. The United States uses them mainly to slow China's progress in AI.
- know-your-customer for computeknow-your-customer for compute, n.: Rules that would make cloud companies check who their customers are before renting them enough computing power to train a large AI model, much as banks must verify who opens an account.
- trusted execution environmenttrusted execution environment, n.: A sealed-off part of a computer chip that runs programs which the rest of the machine, even its operating system, cannot spy on or interfere with.
- confidential computingconfidential computing, n.: Running computations inside a hardware-sealed compartment of a chip, so that data and AI models stay protected even from the cloud company or administrator operating the machine.
- AI acceleratorAI accelerator, n.: Any chip built or tuned to speed up AI — graphics chips, Google's TPUs, phone 'neural engines' and many startup designs. Such chips do the heavy math of training and running models.
Sources
- Aarne, O., Fist, T. & Withers, C. (2024). Secure, Governable Chips: Using On-Chip Mechanisms to Manage National Security Risks from AI & Advanced Computing. Center for a New American Security.
- Kulp, G., Gonzales, D., Smith, E., Heim, L., Puri, P., Vermeer, M. & Winkelman, Z. (2024). Hardware-Enabled Governance Mechanisms: Developing Technical Solutions to Exempt Items Otherwise Classified Under Export Control Classification Numbers 3A090 and 4A090. RAND Corporation.
Mentioned in
export controls on AI chipsexport controls on AI chips, n.: Government rules that limit which countries and companies may buy the most advanced AI chips and the machines used to make them. The United States uses them mainly to slow China's progress in AI., compute governancecompute governance, n.: Using control over the computer chips and data centers that AI depends on — who can buy them, where they go, how they are used — as a way to oversee powerful AI., kill switchkill switch, n.: A way to quickly shut down an AI system, or halt its development, if it starts behaving dangerously — anything from an off button to a company's promise to stop., know-your-customer for computeknow-your-customer for compute, n.: Rules that would make cloud companies check who their customers are before renting them enough computing power to train a large AI model, much as banks must verify who opens an account.