Coinbase and Bithumb customers exposed by separate data leaks
Coinbase and Bithumb users had data exposed after separate incidents; three other firms were also breached.
Google is launching a multi-party computation crypto custody tool that splits signing keys, according to a report.
At its inaugural digital assets symposium in New York today, Google is reportedly introducing Google Multi-Party Compute, a crypto custody system that divides signing keys across separate systems. The tool is currently in a private preview phase with several financial institutions. Alphabet has also begun using the system to protect its own digital assets, according to CoinDesk.
Commercial availability is scheduled for early 2027, the company said, while pricing details have not been disclosed.
The system applies multi-party computation (MPC) to signing keys. The key is split into fragments held by different parties or machines, and a transaction requires a specific number of fragments to be combined to sign.
The complete key is never reconstructed in a single place. A bank, for example, could store one share on its own hardware, a second in Google Cloud, and a third with a competing cloud provider. Institutions can configure thresholds such as two-of-three or five-of-seven. Before any signature, a policy engine operating within Google’s Confidential Space checks the transaction against limits, destination restrictions, and approval rules.
Google asserts that it cannot access key shares in unencrypted form, according to CoinDesk.
The offering addresses an ongoing challenge. Chainalysis reported that 43.8% of all cryptocurrency stolen in 2024 came from compromised private keys. According to DeFiLlama data, roughly 40% of the $16.69 billion lost to hacks by June resulted from key compromises. In a June attack, a hacker stole seven keys from a single developer’s machine during the Humanity Protocol breach, subsequently draining wallets on two networks.
These incidents illustrate the risk of a single point of failure when signing authority is concentrated. In response, the industry has advocated for multi-vendor multisignature configurations that distribute control across multiple hardware and software providers.
Google also positions the product in the context of AI-driven fraud. Chainalysis data shows that scams associated with AI tool vendors averaged $3.2 million per operation in 2025, compared to $719,000 for other scams. The report connects these figures but does not establish that AI was the cause of the difference.
Ethereum researcher Justin Drake has warned that AI might eventually discover mathematical shortcuts capable of compromising wallet signatures. No AI system has demonstrated such an ability. Google’s MPC system is designed to address stolen credentials and unauthorized transfers, not a break in the underlying cryptography.
Quantum computing presents a distinct, longer-term threat. Google researchers estimated that a sufficiently advanced future quantum computer could derive a bitcoin private key from an exposed public key in approximately nine minutes. Roughly 6.9 million bitcoin have exposed public keys. Google said its architecture can adopt post-quantum signatures later without requiring institutions to rebuild their systems, but no timetable has been set.
Google is entering a market dominated by Fireblocks, Fordefi, and BitGo. The company told CoinDesk that its banking relationships and security expertise provide an advantage. It also positions itself as a neutral provider for banks that do not want to rely on competitors.
Financial institutions already use specialized platforms, including BNP Paribas’s custody arrangement with Metaco and multi-institution vaults built with BitGo.
Share to
Disclaimer: this article comes from third-party media and is provided for reference only. It does not constitute investment advice. Crypto and other financial products carry significant price volatility risk, so please make your own decisions carefully.
Coinbase and Bithumb users had data exposed after separate incidents; three other firms were also breached.
AI agents are making payments but spending limits may not prevent costly errors. A new report highlights risks and the need for stronger controls.
Bitwise lifted its 2036 bitcoin price target to $1,599,766 from $1.3 million, citing institutional buying, fixed supply and government debt.
OpenAI says its model disproved the Exponential Time Hypothesis, but Bitcoin and Ethereum cryptography remain unaffected.