Crypto Was the Dress Rehearsal for Taxing AI
On 1 January 2026, the EU's DAC8 directive started requiring crypto-asset service providers to report their users' transactions to tax authorities. No new crypto tax. No attempt to locate a blockchain. Just a reporting duty on whoever faces the customer.
That choice matters beyond crypto. AI raises the questions crypto hit a decade earlier - who captures value, where it's taxed, whether code can pay tax - and regulators reach for the same answers.
Base layer or apps?
In 2016, Joel Monegro argued blockchain protocols would end up worth more than the apps built on them. Reality disagreed. By 2023, Uniswap and OpenSea alone had made nearly $2.8 billion, and protocols struggle to raise fees without pushing users elsewhere.
AI is rerunning that argument between model labs and the products built on top. Open-weight models play a similar role to forks: cheaper alternatives that can cap what the base layer charges.
Nowhere to tax, so tax the edge
Validators can sit anywhere, like AI servers. DAC8 builds on the OECD's Crypto-Asset Reporting Framework, and countries swap the first reports by 30 September 2027.
AI proposals agree. Reuven Avi-Yonah and co-authors argue autonomous AI should be taxed where its users are, since programmers, servers and users rarely share a country.
Wrap the code, tax the power
Wyoming has let DAOs register as LLCs since 1 July 2021, shielding members and giving the law something to grab. Avi-Yonah's team proposes the AI version: separate legal personhood for autonomous AI inside its own corporate shell, taxed at a rate tied to measured harm.
Energy follows the same script. Biden's FY2024 budget proposed a 30% excise on the electricity cost of crypto mining. It was dropped in the 2023 debt-ceiling deal. In August 2026, Senator Ron Wyden proposed a data-center excise. Same logic: tax the physical bottleneck that can't move.
Where the analogy breaks
Crypto tokens have a public price every second and often thin cash flows. AI labs report real revenue but trade privately. One is price without fundamentals. The other is fundamentals without a price.
Transparency runs backwards too. A blockchain is a public ledger. A model is a black box. Crypto's tax problem was mostly reporting gains. AI's is valuing the asset itself.
The playbook transfers: report at the edge, tax the hardware, wrap the code. It doesn't solve what crypto never faced - pricing an asset nobody can see inside. Who builds that tool first, the auditors or the labs?
Sources: European Commission on DAC8; Decentralised.co (July 2023); Regulation Tomorrow (May 2021); American Public Power Association (Mar 2023); Avi-Yonah et al., Marquette Law Review (2025); Patel & Gordon, Brookings (Aug 2026).