CryptoSlate
A soundness bug inside Orchard, Zcash's newest zero-knowledge shielded pool, allowed the proof system to accept a transaction it should have rejected — the exact mechanism that, in theory, lets a counterfeiter mint unlimited ZEC off-chain. Security researcher Taylor Hornby at Shielded Labs found the flaw on May 29 during a targeted review he ran with Anthropic's Opus 4.8, released the day prior, paired with a custom AI harness. Zcash executed an emergency soft fork at block 3,363,426 on June 2 at 02:00 UTC to disable Orchard actions, then completed the NU6.2 hard fork on June 3 at 00:05 EDT at block 3,364,600, replacing the circuit and restoring full functionality in under five days from discovery.
Why it matters
Orchard is not a smart contract or a DeFi vault — it is the cryptographic core of Zcash's private transaction system, and a soundness bug at that layer is a money-layer failure, not an app-layer one. Zcash's official position is that turnstile accounting detected no unauthorized value creation and the 21 million ZEC cap remains intact. Shielded Labs takes a harder line: Orchard's privacy guarantees make it cryptographically difficult to prove the negative, and the firm is pushing a follow-on upgrade that would route existing Orchard coins back through turnstile accounting so anyone can verify supply integrity on-chain. The pattern repeats — Octane's AI found a Nethermind execution-client bug in February 2026 that could have halted local block production for roughly 38% of Ethereum validators — and an arXiv paper from January reported a 63% success rate for AI-agent exploit generation against smart contract benchmarks. AI-assisted research is now reaching consensus clients, ZK circuits, and supply rules, a threat layer most base-layer systems were not designed against.
Frequently asked questions
-
What was the Zcash Orchard soundness bug?
A flaw in the zero-knowledge proof circuit powering Orchard, Zcash's newest shielded pool, that allowed the proof system to accept a transaction it should have rejected — the mechanism that, on mainnet, could have generated unlimited counterfeit ZEC without detection.
-
How did the researcher find the flaw?
Taylor Hornby of Shielded Labs found it on May 29 during a targeted review conducted with Anthropic's Opus 4.8, released the day prior, alongside a custom AI harness and prompts that produced a working local exploit in a regtest environment.
-
How did Zcash respond to the bug?
Zcash executed an emergency soft fork at block 3,363,426 on June 2 at 02:00 UTC to disable Orchard actions, then completed the NU6.2 hard fork on June 3 at 00:05 EDT at block 3,364,600, replacing the circuit and restoring full Orchard functionality in under five days from discovery.
-
Was the 21 million ZEC supply cap actually breached?
Zcash says no — turnstile analysis found no evidence of unauthorized value creation. Shielded Labs argues the privacy properties of Orchard make that negative cryptographically hard to prove and is pushing a follow-on upgrade to route existing Orchard coins through turnstile accounting for on-chain verification.
-
How does this connect to the broader AI-exploit trend?
Octane's AI surfaced a Nethermind execution-client bug in February 2026 that could have halted local block production for roughly 38% of Ethereum validators, and a January 2026 arXiv paper reported a 63% success rate for AI-agent exploit generation on smart contract benchmarks. AI-assisted research is now reaching…
WuBlockchain
Crypto Rank News
Lookonchain
TheBlock
CoinTelegraph
Glassnode
CoinDesk