Crypto Exchanges Seek AI Models Capable of Autonomous Smart Contract Hacking
Simulated damage worth millions of dollars resulted from AI agents using the Claude Mythos Preview model to autonomously hack smart contracts. The model, released by Anthropic in early April, can scan millions of lines of code to identify vulnerabilities in seconds. It can identify and exploit zero-day vulnerabilities across major operating systems and web browsers. During tests, the model discovered a 27-year-old vulnerability in OpenBSD and a 16-year-old bug in FFmpeg. Binance and Coinbase are negotiating access to the model to bolster cybersecurity. Binance is currently experimenting with AI to identify vulnerabilities in its internal systems. Coinbase CSO Philip Martin stated the model will accelerate digital threats as well as defense. Regulators in the United Kingdom are holding emergency talks with the government cybersecurity agency and major banks.
More Briefs
BlackRock leverages private market acquisitions to outpace S&P 500 decline
Apr 16The top 1% of Social Security retirees receive checks averaging $4,140 monthly
Apr 16Cypherpunk Technologies now owns 1.82% of the Zcash network
Apr 16S&P 500 Gains 10% After Geopolitical Shock as Energy Intensity Drops