US Government Lifts Ban on Anthropic's Mythos 5 as Frontier AI Enters 'Vetted Release' Era
In a major de-escalation of tensions between Washington and the AI industry, the Trump administration has granted Anthropic permission to redeploy its 'Mythos 5' and 'Fable 5' models to roughly 100 trusted U.S. institutions. The models were abruptly disabled two weeks ago by a Commerce Department export order due to national security and foreign access concerns. Simultaneously, OpenAI has begun a restricted rollout of its new GPT-5.6 series (Sol, Terra, and Luna), which remains accessible only to government-approved partners for cybersecurity review.
OpenAI and Broadcom unveil 'Jalapeño,' a custom silicon solution purpose-built for high-speed model inference.
Meta announces plans to replace 90% of its content review staff with specialized AI agents, citing increased accuracy and efficiency.
Google DeepMind publishes 'From AGI to ASI,' a roadmap identifying four algorithmic and compute-based pathways to artificial superintelligence.
Kazakhstan joins the 'Pax Silica' partnership, an international coalition focused on securing AI infrastructure and supply chains.
Anthropic officially retires Claude 3.5 Sonnet as the new 'Claude Code' and 'Cowork' agentic tools take over professional workflows.
Did you know?
OpenAI's new custom inference chip, 'Jalapeño,' developed with Broadcom, reportedly reduces LLM serving costs by 50% compared to standard high-end GPUs.
Quiz
Which specific OpenAI model series, released under government restrictions on June 26, 2026, includes versions optimized for different reasoning tiers?