Key Highlights
- The cryptocurrency exchange Coinbase is negotiating with Anthropic for access to the exclusive Claude Mythos Preview AI platform.
- This initiative targets enhanced cybersecurity defenses and streamlined operations across the platform.
- In early April 2026, Anthropic unveiled Project Glasswing, offering controlled access to Mythos for select collaborators.
- Testing phases revealed the model’s ability to detect thousands of previously undiscovered zero-day security flaws, including a vulnerability in OpenBSD dating back 27 years.
- Initial Glasswing collaborators comprise AWS, Apple, Google, JPMorgan Chase, Microsoft, and Palo Alto Networks.
The cryptocurrency platform Coinbase is pursuing discussions with Anthropic to secure access to Claude Mythos Preview, an exclusive artificial intelligence system designed for sophisticated cybersecurity applications, as reported by The Information.
This strategic move emerges as digital asset platforms encounter escalating threats from artificial intelligence-enhanced attack vectors that traditional defenses struggle to counter.
Anthropic introduced Project Glasswing during the first week of April 2026. This specialized program provides a select consortium of partners with controlled access to Mythos exclusively for defensive cybersecurity applications.
The initial partnership roster features Amazon Web Services, Apple, Google, JPMorgan Chase, Microsoft, and Palo Alto Networks. An additional 40-plus organizations responsible for maintaining essential software infrastructure have also gained entry.
As part of this initiative, Anthropic allocated $100 million in computational resources and contributed $4 million toward open-source security communities.
Throughout the testing period, Mythos uncovered thousands of zero-day vulnerabilities that had never been documented before. Notable discoveries included a security flaw in OpenBSD that had existed undetected for 27 years and a 16-year-old defect in FFmpeg.
These discoveries have drawn significant interest from major institutions like Coinbase, which currently deploys Claude AI technology for customer service operations spanning over 100 geographic markets.
The Strategic Rationale Behind Coinbase’s Interest
Coinbase experienced a significant internal security compromise in 2025. Criminal elements successfully bribed international customer support personnel, resulting in the exposure of confidential information affecting approximately 70,000 platform users.
The exchange rejected a $20 million extortion demand and instead offered an equivalent reward for intelligence leading to the perpetrators’ capture.
Research conducted by Anthropic has demonstrated that autonomous AI systems can independently identify and exploit vulnerabilities in smart contracts, producing simulated theft scenarios worth millions of dollars.
Such findings transform Mythos access from an optional enhancement into a critical defensive resource for platforms managing substantial financial assets.
Pricing Structure and Accessibility Restrictions
Mythos will remain unavailable for general commercial deployment. Anthropic intends to integrate its advanced features into subsequent Claude versions equipped with enhanced protective measures.
Following the preview phase, the pricing model will be established at $25 per million input tokens and $125 per million output tokens.
Confirmation remains pending regarding whether Coinbase will obtain official Glasswing partnership designation or expanded model access privileges.
Negotiations continue between both entities, with neither company having issued formal statements confirming any definitive arrangement.
