Anthropic's release of Mythos marks a watershed moment for AI development access. The model, previously restricted to cleared defense contractors, now reaches the broader developer community. This collapse of gatekeeping between Pentagon-grade and public-facing AI reshapes the competitive landscape instantly.
DeepMind's Demis Hassabis sharpened his AGI timeline. He moved from the vague "five to ten years" estimate to "a real possibility by 2029," tying this forecast directly to AlphaProof Nexus solving nine open Erdős problems. The mathematical breakthroughs demonstrate concrete progress toward systems capable of original research.
Critical security gaps emerged across AI infrastructure. A zero-day vulnerability hit Starlette, affecting roughly one million AI agents in production. CrowdStrike coordinated a takedown of Glassworm, a developer botnet operating across four command-and-control channels, signaling rising operational security concerns as AI systems proliferate.
Geopolitical stakes intensified on multiple fronts. BNP Paribas and Mistral formalized a sovereign-AI security partnership, positioning Europe to reduce dependence on U.S. and Chinese AI stacks. Beijing responded by freezing overseas travel for top AI researchers at Alibaba and DeepSeek, a hard signal that China views AI talent as strategically critical.
The economics of AI displacement became measurable. Uber exhausted its annual AI token budget by April, revealing the real cost of deploying frontier models at scale. This spending pattern exposes the gap between marketing narratives about AI efficiency and actual compute requirements.
The week crystallizes three realities. First, frontier AI access is democratizing rapidly, collapsing the technical moats that once protected government-cleared models. Second, AGI timelines are contracting as architectural breakthroughs stack. Third,