Binance and Coinbase Seek Access to Anthropic's Claude Mythos AI Model Capable of Hacking Smart Contracts

Major crypto exchanges Binance and Coinbase are in active negotiations with Anthropic to gain access to the Claude Mythos Preview model, according to The Information. The AI was released in early April but withheld from public access due to its extraordinary ability to detect and exploit software vulnerabilities.

What Is Claude Mythos Preview

Anthropic launched Claude Mythos Preview in early April 2026 but restricted access to members of Project Glasswing. The approved testing group includes AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, and several other firms.

The restricted rollout stems from the model's unprecedented capabilities. According to developers, Mythos can scan millions of lines of code and identify vulnerabilities within seconds. During testing, the AI discovered a 27-year-old vulnerability in OpenBSD and a 16-year-old bug in FFmpeg — flaws that had gone undetected for decades.

Internal research by Anthropic revealed an even more alarming dimension: AI agents powered by Mythos can autonomously hack smart contracts, causing simulated damages in the millions of dollars. Traditional security auditing takes weeks by comparison, while the new model identifies effective attack vectors in mere seconds.

Why This Matters

An AI capable of autonomously discovering and exploiting smart contract vulnerabilities fundamentally reshapes the cybersecurity landscape for the crypto industry. Exchanges custody billions of dollars in assets, making a tool of this caliber both a powerful shield and a potential weapon. For Binance and Coinbase, securing access to Mythos is a matter of competitive advantage and operational security at a time when vulnerability detection speed is reaching an entirely new level.

How the Exchanges Are Responding

Coinbase Chief Security Officer Philip Martin confirmed the negotiations with Anthropic, stating that Mythos and future models of this class will enable far deeper testing of software and systems at scale. He emphasized that this will accelerate both digital threats and digital defense capabilities.

Binance Head of Security Jimmy Su noted that the exchange is exploring how AI advancements can strengthen cybersecurity while acknowledging the new risks they create. He said Binance is already experimenting with AI tools to find vulnerabilities in internal systems more rapidly.

Market and Regulatory Reaction

Clear Street analyst Owen Lau suggested that the threat from Mythos remains unclear at this stage. He believes it will create a negative narrative for crypto exchanges, but only temporarily. In his view, these platforms will be among the first to build defenses against such AI agents over the long term.

Regulators worldwide have already taken notice of Mythos' capabilities. The United Kingdom has initiated emergency consultations with its government cybersecurity agency and the country's largest banks. The scale of concern signals that the emergence of such a powerful model is perceived as a systemic risk extending well beyond the crypto industry.