AI Agents Just Slashed the Cost of a Quantum Attack on Bitcoin

Researchers working with AI coding agents reduced a resource benchmark for a key step in a potential quantum attack on the cryptography protecting Bitcoin and Ethereum by 86%, according to a paper posted Wednesday.

In other words, they reduced the computing resources needed for uncovering a wallet’s private key, which could let an attacker steal its funds. When quantum computers powerful enough to carry out a full attack will arrive—a milestone known as “Q-Day”—remains uncertain.

“Although its timing remains uncertain, migration away from vulnerable cryptography is already under way,” the researchers wrote. “NIST has standardized post-quantum replacements, and the initial public draft of NIST IR 8547 proposes deprecating classical public-key algorithms at the 112-bit security level after 2030 and disallowing them after 2035.”

The findings came from ECDSA.Fail, an open competition launched by Eigen Labs in late May. More than 100 participants worked on circuits for secp256k1, the elliptic curve used to secure transaction signatures on Bitcoin and Ethereum. The paper’s authors include researchers affiliated with Theta Labs, MultiVM Labs, Eigen Labs, Trail of Bits, StarkWare, and the Ethereum Foundation.

Researchers designed and tested a quantum circuit that performs a calculation needed to crack Bitcoin’s cryptography, cutting its resource score by 86%. The tests verified the circuit’s calculations; they did not crack a Bitcoin private key.

The challenge measured how much quantum memory and computational work each circuit needed, multiplying its logical qubits by its Toffoli gates, a costly type of quantum operation invented by Tommaso Toffoli in 1980. A lower score means the calculation requires fewer combined resources, potentially making a future attack easier to carry out.

By July 26, participants had cut that score by 86%, from 10.75 billion to 1.496 billion. The leading design used 1,151 logical qubits and about 1.3 million Toffoli gates. According to the study, the score was roughly half of Google Quantum AI’s March benchmark, though differences in testing and counting methods prevent a direct comparison.

“Beyond the resulting circuits, ECDSA.Fail provides a case study of Open Autoresearch: a verifier-gated research process in which human participants and AI agents iteratively generate, implement, test, and share candidate improvements against a common measurable objective,” the researchers wrote.

The report is a part of growing efforts by crypto companies to invest in research to defend against quantum attacks.

In July, Galaxy Digital committed up to $5 million to that work, while nine firms—including BlackRock, Coinbase, and Strategy—pledged a combined $15 million over three years for broader Bitcoin security research, including quantum defenses.