Claude Mythos on Bedrock (gated)

Amazon Web Services has put Anthropic’s Claude Mythos Preview into Amazon Bedrock, but only as a gated research preview for cybersecurity work. (aws.amazon.com) Amazon announced the Bedrock release on April 7, 2026, as part of Project Glasswing. Amazon’s Bedrock documentation says access is being prioritized for defensive cybersecurity use cases rather than general commercial deployment. (aws.amazon.com) (docs.aws.amazon.com) Bedrock is Amazon’s managed service for using outside and in-house foundation models through one application programming interface, so adding Mythos there gives selected customers a standard way to test it inside existing Amazon Web Services setups. Amazon says Bedrock is built for production-scale generative artificial intelligence applications and agents. (aws.amazon.com 1) (aws.amazon.com 2) Anthropic is framing the release around software defense, not chatbot use. On its Project Glasswing page, the company says the program is aimed at securing critical software and gives defenders early access to Mythos Preview to find and patch vulnerabilities. (anthropic.com 1) (anthropic.com 2) That focus follows Anthropic’s own testing claims about the model’s security abilities. In a technical write-up, Anthropic said Mythos Preview could identify and exploit zero-day vulnerabilities across every major operating system and every major web browser during internal testing, and said more than 99 percent of the vulnerabilities it found had not yet been patched. (red.anthropic.com) Anthropic says Project Glasswing launched with Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks as partners. It also says it extended access to more than 40 additional organizations that build or maintain critical software infrastructure. (anthropic.com 1) (anthropic.com 2) Amazon’s security blog says organizations in the program are using Mythos Preview to scan first-party and open-source systems for flaws before attackers do. That puts Bedrock in the middle of a model rollout designed less like a normal product launch and more like a controlled security program. (aws.amazon.com) Anthropic has also said it does not plan to make Claude Mythos Preview generally available. The company’s red-team site says the longer-term goal is to let users safely deploy Mythos-class models at scale, but only after building the security controls around them. (red.anthropic.com) So the immediate change is narrow but concrete: selected partners can now reach Mythos through Amazon Bedrock instead of a separate bespoke channel. The broader message from Amazon and Anthropic is that the model is being released first where it can be watched most closely. (aws.amazon.com) (docs.aws.amazon.com)