Anthropic won't release its new model

Anthropic

Anthropic's new Claude Mythos Preview AI model can find software vulnerabilities at elite human hacker levels, prompting a defensive cybersecurity initiative called Project Glasswing with tech giants including Amazon, Apple, Google, and Microsoft. The company is committing $100M+ to help secure critical infrastructure while restricting general access due to the model's dual-use risks.

Anthropic Blog

Anthropic's unreleased Claude Mythos model showcases powerful autonomous cybersecurity exploitation abilities, highlighting the growing tension between AI capability advancement and responsible deployment practices.

Anthropic

Anthropic published a detailed 244-page system card for Claude Mythos, marking a major advancement in AI transparency and safety documentation with focus on cybersecurity capabilities.

Anthropic

Anthropic's new Claude Mythos model and Project Glasswing initiative aim to secure critical software by detecting vulnerabilities at near-expert human levels, demonstrating frontier AI's emerging role in enterprise cybersecurity.

Z.ai

Z.ai's GLM-5.1 achieves state-of-the-art on software engineering benchmarks and introduces breakthrough long-horizon optimization capabilities, sustaining improvement over hundreds of iterations where previous models plateau. The open-source model demonstrates 6× gains on vector database tasks and builds complete applications autonomously through iterative refinement.