Anthropic AI Updates: April 21, 2026

1. Amazon Adds $5B to Anthropic in Exchange for $100B+ AWS Spend Commitment

Anthropic. Amazon committed an additional $5 billion to Anthropic with up to $20 billion to follow, bringing total investment to roughly $13 billion at a $350 billion pre-money valuation. In return, Anthropic agreed to spend more than $100 billion on AWS over 10 years and will receive up to 5 GW of compute, including access to Trainium2 through Trainium4 silicon. The deal mirrors Amazon’s recent $50B OpenAI commitment and re-anchors Anthropic in AWS even as it maintains parallel partnerships with Google and Broadcom. Source

2. NSA Reportedly Among 40 Organizations Using Mythos Preview Despite Pentagon Blacklist

Anthropic. Axios reported and TechCrunch corroborated that the NSA is among about 40 organizations granted access to Mythos Preview, the frontier cybersecurity model Anthropic withheld from public release citing offensive-cyber risks. The agency is reportedly using the model to scan its own environments for exploitable vulnerabilities. The disclosure is notable because the Department of Defense, which oversees the NSA, formally labeled Anthropic a “supply chain risk” in February. Source

3. Built with Opus 4.6 Hackathon Awards $100K Across Six Projects

Anthropic. Anthropic announced the winners of its inaugural “Built with Opus 4.6” virtual hackathon, distributing $100,000 in prizes after seeding 500 participants with $500 in API credits each. First place went to CrossBeam, a housing-permit automation tool by Mike Brown, with other winners spanning education, healthcare, road infrastructure, and AI music collaboration. Anthropic noted four of the five main winners were non-professional developers and previewed an Opus 4.7 hackathon. Source