Anthropic releases Pentagon-grade Mythos AI to the public

Anthropic shocked the AI world by releasing Mythos, a model previously deployed inside NSA and Pentagon supply-chain operations, to the public. Now accessible via a standard API key, Mythos collapses the capability gap between sovereign-grade and developer-grade AI, forcing OpenAI and xAI to either match or fall behind the public frontier. This marks the first time national-security-grade AI capabilities have crossed openly into developer access.

In related frontier news, DeepMind CEO Demis Hassabis compressed his AGI timeline from a five-to-ten-year window to a specific date—2029. He tied this directly to AlphaProof Nexus, which solved nine open Erdős problems and 44 OEIS conjectures for a few hundred dollars in compute. The claim is under review by Cambridge mathematicians, but it signals a dramatic acceleration in AI reasoning capabilities.

The AI supply chain faced simultaneous crises: a critical zero-day (CVE-2026-48710) in Starlette, affecting 400,000+ GitHub dependents including vLLM, LiteLLM, and Ray Serve, allowed auth bypass via Host-header injection. Automated exploit tooling is already in the wild. Meanwhile, CrowdStrike, Google, and Shadowserver coordinated a takedown of the Glassworm botnet, which had compromised 300+ repos across npm, PyPI, and GitHub since 2025 using four C2 channels including Solana blockchain and Google Calendar titles.

Geopolitical shifts continued: BNP Paribas formalized a sovereign-AI security partnership with Mistral, while Beijing froze overseas travel for top AI engineers at Alibaba and DeepSeek. On the workforce front, Uber burned its full-year AI token budget by April, ClickUp restructured to 1,000 humans alongside 3,000 internal agents, and Sam Altman publicly reversed his earlier white-collar-apocalypse prediction.