Two banks, three regulators, one 2-of-2 key: what the NEAR post-quantum pilot actually tests

A 2-of-2 threshold has no spare capacity. Both parties hold a share, both must cooperate to sign, and neither ever possesses the full private key ^[7], which means the number of shares that can go missing before signing stops is zero ^[21]. Neither published account identifies who holds the second share alongside the bank ^[23]. For a custody operator that matters more than the algorithm choice, because the practical failure mode of a quantum-safe wallet in production tends to be a counterparty that cannot co-sign rather than a broken curve. A testnet where no real assets are exposed ^[10] is a sensible place to confirm that wallet generation and transfers validate ^[14]. It is not where availability under stress gets priced.

The two reports agree on the calendar and diverge on the substance. Cointelegraph dates the announcement to a Monday ^[19], and August 24, 2026 was a Monday ^[20]. But Crypto Briefing describes the supervisors as participating in governance discussions to align outcomes with existing regulatory frameworks ^[12], while Cointelegraph has them observing the first phase and contributing to a governance workstream later, with participation levels differing between institutions ^[13]. Those are different products. One is co-designed supervision; the other is three agencies watching a vendor demo with an option to engage. Until the whitepaper lands ^[15], the narrower reading is the safer one.

The audit sequencing points the same way. Safeheron, the custody technology firm supplying the cryptographic backbone ^[5], co-announced the pilot with RFI ^[5], and the underlying protocols are to be open-sourced eventually rather than at launch ^[15]. So the first artifact anyone outside the pilot can inspect is a document written by the participants, not code. ML-DSA-65 itself is the least contestable part of the stack, having been standardized by NIST in FIPS 204 at the medium security level of that family ^[8]. The integration around it is the part that needs eyes.

Worth noting that the stated urgency and the chosen tool are not the same problem. Harvest now, decrypt later describes intercepting ciphertext today and decrypting it once machines are strong enough ^[11]. A digital signature standard ^[6] does not address that; what it addresses on a public chain is Shor's algorithm recovering keys from the elliptic-curve cryptography securing most wallets today ^[9].

The clock explains the timing better than the threat narrative does. The Hong Kong Monetary Authority wants its banking sector fully prepared for quantum-related security risks by 2030 ^[16], which is about 41 months from this pilot's launch ^[22]. A 2025 BIS paper told institutions to start coordinated, phased migrations ^[17], and MAS with the Association of Banks in Singapore folded quantum threats into a cyber taskforce ^[18]. Supervisors need something to point at when they ask banks what phase one looked like. Two banks on a testnet is now that thing.