From AI Proximity to AI Causation: An Evidence-Calibrated Reconstruction of the 2026 Binance Breach
Year: 2026
Type: preprint
Abstract: The October 29, 2026 Binance incident combined developer-targeted social engineering, endpoint compromise, an internal AI tool, deployment authority, wallet-signing infrastructure, and EIP-7702 delegation in an unauthorized transfer event reported at approximately USD 6.34 billion. This study separates public-chain observations, first-party reporting, official attribution, third-party inference, and unresolved unknowns. Direct Ethereum reproduction recovered Binance 14 as signer of 50 descending authorization tuples, reconstructed the same-block nonce change that made exactly one tuple effective, verified delegation and revocation, and showed that delegated ownership authority enabled two downstream smart-account calls transferring 823 million USDC. The public evidence supports AI proximity and a company-reported instrumental role, but not the claim that AI-generated code, autonomous decision-making, or “vibe coding” caused the breach.
Author: Alkaslassy, Nadav
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225754625
Cites:
Cited by:
Related to: