Anthropic's release of Mythos marks a watershed moment in AI accessibility. The model, previously restricted to cleared defense contractors, is now available to the general developer community. This collapse of the barrier between classified frontier AI and commercial models compresses a significant gap that separated government-grade systems from what developers could access.
Demis Hassabis updated DeepMind's AGI timeline to 2029, moving from the previous five-to-ten-year range. He anchored this prediction to AlphaProof Nexus, which solved nine open Erdős problems at minimal computational cost. This represents a concrete performance milestone rather than speculation.
Security concerns surfaced across the ecosystem. Starlette vulnerabilities exposed critical infrastructure affecting millions of AI agents. CrowdStrike orchestrated a takedown of the Glassworm developer botnet across four command-and-control channels, indicating organized malicious activity in the AI development space.
Geopolitical jostling intensified. BNP Paribas formalized a sovereign-AI security partnership with Mistral, signaling institutional confidence in European alternatives to US systems. Beijing restricted overseas travel for top AI engineers at Alibaba and DeepSeek, a move that typically precedes intensified domestic development efforts or prevents knowledge transfer.
The economic reality of AI deployment became harder to ignore. Uber exhausted its full annual AI token budget by April, exposing the gap between AI capability and sustainable commercial deployment. The company's burn rate suggests that current AI systems demand resources that outpace practical revenue generation at scale.
These developments converge on a simple fact: frontier AI moved from classified labs into production systems this quarter. The security gaps are now public. The timeline to AGI compressed. The cost of deployment became visible. The geopolitical competition turned explicit. What was once theoretical debate about AI's trajectory now involves budgets