Key Highlights
- The cryptocurrency exchange is currently negotiating with Anthropic for access to the exclusive Claude Mythos Preview AI system.
- This initiative focuses on enhancing defensive capabilities and boosting operational performance across the platform.
- Anthropic unveiled Project Glasswing in April 2026, providing restricted Mythos access to carefully chosen organizations.
- The AI system discovered numerous previously unknown security flaws during trials, including a vulnerability in OpenBSD dating back 27 years.
- Initial Glasswing participants include major tech and financial institutions such as AWS, Apple, Google, JPMorgan Chase, Microsoft, and Palo Alto Networks.
The leading cryptocurrency platform is pursuing an agreement with Anthropic to secure access to Claude Mythos Preview, an exclusive AI system designed specifically for sophisticated cybersecurity applications, The Information has revealed.
This development emerges amid mounting challenges faced by digital asset platforms dealing with increasingly sophisticated AI-driven cyber attacks that traditional security measures struggle to counter.
Anthropic introduced Project Glasswing in April 2026, establishing a selective consortium that grants Mythos access exclusively for protective security applications.
The founding consortium members encompass Amazon Web Services, Apple, Google, JPMorgan Chase, Microsoft, and Palo Alto Networks. Additionally, over 40 organizations responsible for essential software infrastructure have been granted participation.
Anthropic allocated $100 million in computational resources and dedicated $4 million toward open-source security communities through this program.
Throughout initial trials, Mythos uncovered thousands of security vulnerabilities that had never been documented before. Among these discoveries were a 27-year-old security gap in OpenBSD and a defect in FFmpeg that had existed for 16 years.
These revelations have drawn significant interest from prominent platforms like Coinbase, which currently leverages Claude technology for customer assistance operations spanning more than 100 geographic markets.
The Strategic Importance for Coinbase
Coinbase experienced a significant internal security incident in 2025 when international customer service representatives were compromised through bribery schemes by criminal organizations, resulting in the exposure of sensitive information belonging to approximately 70,000 platform users.
The cryptocurrency exchange declined to pay a $20 million extortion demand and instead established an equivalent reward for intelligence that could lead to the apprehension of those responsible.
Research conducted by Anthropic has demonstrated that autonomous AI systems possess the capability to identify and exploit weaknesses in blockchain smart contracts, potentially facilitating the theft of millions in simulated scenarios.
Such capabilities position Mythos access as a critical defensive asset rather than an optional enhancement for platforms managing substantial financial assets.
Access Requirements and Pricing Structure
Mythos will remain unavailable for general public use. Anthropic intends to integrate its advanced features into subsequent Claude versions equipped with enhanced security protocols.
The pricing structure following the preview period is established at $25 per million input tokens and $125 per million output tokens.
Confirmation regarding whether Coinbase will obtain official Glasswing partnership designation or expanded model access remains pending.
Discussions between the two organizations continue, with neither party having issued an official statement confirming any finalized arrangement.
