OECD Crypto-Asset Reporting Framework Enters Implementa...

The OECD's Crypto-Asset Reporting Framework is no longer a proposal or a commitment letter. As of 1 January 2026, exchanges, brokers, and custodial wallet providers in 52 jurisdictions began collecting identity and transaction data on their crypto users. First automatic exchanges between tax authorities are scheduled for 2027. A second wave of 23 jurisdictions (including the UAE, Singapore, Hong Kong, and Australia) starts data collection in 2027 with exchanges in 2028. The United States, characteristically, is on its own timeline: data collection from 2028, exchanges in 2029.

What gets reported, and by whom

CARF requires reporting entities to collect and transmit data on purchases, sales, exchanges, transfers (including to cold wallets), and payments made using crypto assets. The data package for each user includes full name, address, date of birth, tax identification number, and country of tax residence. Reporting entities include centralised exchanges, brokers, custodial wallet providers, crypto ATM operators, and certain DeFi platforms where an identifiable operator exists. Traditional financial institutions that have added crypto trading or custody services also fall within scope.

The 75 committed jurisdictions cover most major financial centres. Wave 1's 52 jurisdictions include the entire EU (via DAC8, the EU's own binding implementation), the UK, Canada, Japan, South Korea, and Brazil. Wave 2 adds jurisdictions that committed later or needed more implementation time.

The EU's parallel track: DAC8

EU member states are implementing CARF through DAC8, which is legally binding rather than voluntary. The transposition deadline was 31 December 2025, with data collection effective from 1 January 2026 and first reporting to tax authorities due by September 2027. DAC8 goes further than base CARF in several respects: penalties for non-compliant reporting entities range from EUR 20,000 to EUR 500,000, platforms must send GDPR-mandated notifications to users about what data has been shared, and exchanges can block customer accounts if users fail to provide required tax certifications.

Gaps and scepticism

Seventy-five countries sounds like near-universal coverage. It isn't. Notable non-participants include Argentina, El Salvador, Georgia, India, and Vietnam. Some of these are significant crypto markets. India, which imposed a 30% flat tax on crypto gains in 2022, has not committed to CARF exchanges despite being an active member of the G20 that mandated the framework's development. El Salvador, which adopted Bitcoin as legal tender, is unsurprisingly absent.

Enforcement readiness is the bigger question. The OECD released XML schemas and IT format specifications in October 2024, giving jurisdictions just over a year to build the technical infrastructure for receiving, processing, and matching millions of transaction records against taxpayer files. How many of the 52 Wave 1 jurisdictions will actually exchange usable data in 2027? The Common Reporting Standard for traditional financial accounts, which launched in 2017, took several years before data quality reached levels that generated meaningful compliance outcomes. CARF will likely follow a similar curve.

The US delay to 2029 creates a three-year window during which American crypto users' data flows to no foreign tax authority through CARF channels. US-based exchanges will not report under CARF until 2028 data collection begins, though domestic reporting obligations under the Infrastructure Investment and Jobs Act's broker provisions (Section 6045) are being phased in on a parallel track.

For crypto holders who assumed that using a foreign exchange or holding assets in self-custody placed them beyond tax authority visibility, CARF changes the calculus. Self-custody wallets are not directly captured (there is no reporting entity), but transfers from a reporting entity to a self-custody wallet are reported, creating a trail that tax authorities can follow. The era of crypto as a practical tool for tax opacity is closing, unevenly and slowly, but closing.