More than thirty bitcoin and cryptocurrency companies have appealed to leading artificial intelligence laboratories to provide open source developers with early access to the most powerful models. The corresponding open letter was organized by the Bitcoin Policy Institute, and among the signatories were such well-known organizations as Coinbase, BitGo, Blockstream and ARK Invest. The authors of the appeal stated that the creators of critical financial infrastructure are now working with less powerful tools than potential attackers. The protective limitations of public neural networks often prevent legitimate vulnerability searches, while new cyber capabilities are rapidly spreading through open-weight models and specialized hacking tools. Advanced artificial intelligence is fundamentally changing the economics of security research, and defenders urgently need access to these technologies.

According to the signatories, early access to advanced models would make it much faster to find and fix vulnerabilities in software that protects trillions of dollars of digital assets. The bitcoin network independently secures funds worth more than one trillion dollars, so any vulnerability in the associated infrastructure can endanger users' capital. Crypto companies have been called upon to create permanent trusted access programs, including the allocation of sufficient computing resources and secure environments for private code analysis. Recent incidents have already demonstrated the relevance of such a step. The developers of BTCPay Server have revealed a critical vulnerability that hackers used to attack Lightning Network nodes. A special group of researchers has already started using neural networks to audit code bases and has filed thousands of reports of potential problems.

The scale of the recent attack on Coldcard hardware wallets has clearly shown how serious the consequences of errors in the cryptographic infrastructure can be. The attackers stole more than one thousand seven hundred and nineteen bitcoins worth about one hundred and eleven million dollars due to a disabled hardware random number generator. The problem of access to advanced tools is especially relevant against the background of a series of devastating attacks on the cryptocurrency infrastructure in the year 2000. The Ostium decentralized protocol lost about twenty-four million dollars due to a vulnerability in the price oracle update mechanism. Similar scenarios were observed when hacking other large platforms, where hackers stole millions of dollars by manipulating the accounting logic of smart contracts using large-scale flash loans. It is precisely such complex vulnerabilities that can become the main area where advanced artificial intelligence models can radically strengthen the protection of the crypto infrastructure. Neural networks can analyze huge code bases and help developers identify atypical exploitation scenarios before they are exploited by intruders.