AI industry crossed a new threshold this week when Google, Microsoft, and Elon Musk’s xAI agreed to give the United States government pre-release access to their most powerful AI models, joining OpenAI and Anthropic in a landmark arrangement with the US Commerce Department’s Center for AI Standards and Innovation.
The agreement, announced Tuesday by the agency, allows government officials to assess AI capabilities and security risks before any new model goes public. OpenAI and Anthropic, which had existing partnerships with the center, simultaneously renegotiated their agreements to align with priorities set out in President Trump’s AI Action Plan.
The move reflects a new era in AI governance, one where the most powerful AI systems are treated not as commercial products alone, but as national security infrastructure. Since 2024, the government office has accessed and evaluated models from OpenAI and Anthropic before their public release. The expansion to cover Google’s Gemini platform, Microsoft’s AI stack, and xAI’s Grok models dramatically broadens the scope of federal oversight.
The announcement lands as the AI industry undergoes its most turbulent period of consolidation and acceleration. OpenAI recently raised $122 billion at a $852 billion valuation, anchored by Amazon, Nvidia, SoftBank, and Microsoft. Anthropic secured an additional $40 billion from Google and $5 billion from Amazon. The sums involved are staggering by any historical standard.
Two frontier AI models cleared a 32-step end-to-end cyberattack simulation range within the same month in recent testing. Anthropic’s Claude Mythos Preview completed it first, with OpenAI’s GPT-5.5 following three weeks later. Britain’s AI Security Institute concluded from these tests that frontier cyber-offense capability is doubling every four months, a finding that has alarmed defense officials on both sides of the Atlantic.
OpenAI has also introduced three new real-time audio models this month, including GPT-Realtime-2, designed for conversational AI agents that handle live voice interactions; GPT-Realtime-Translate, capable of multilingual translation across more than 70 languages; and GPT-Realtime-Whisper, built for live transcription and captioning. Early testing partners include Zillow, and companies across customer service, education, and healthcare are expected to follow.
Meanwhile, Amazon Web Services has built a payments infrastructure for AI agents using Coinbase’s x402 protocol and USDC stablecoins, enabling AI systems to pay independently for APIs, data feeds, and online services without human intervention. The initiative points toward a near-future where AI agents conduct commercial transactions autonomously.
On the geopolitical front, China’s National Development and Reform Commission formally blocked Meta’s $2 billion acquisition of Chinese AI startup Manus, deepening the technology decoupling between Washington and Beijing. The move underscores how AI has become a central battleground in great-power competition, with both governments treating model access and chip supply chains as matters of national security.
Read More: U.S. H-1B Visa Crackdown and AI Boom Collide in 2026, Splitting the American Tech Economy in Two
The Microsoft-OpenAI relationship also shifted this month. The two companies reset their original exclusive partnership to a non-exclusive arrangement, with Microsoft remaining the primary cloud partner and retaining an IP license through 2032. That restructuring followed months of tension as OpenAI pursued broader commercial partnerships with Amazon and others.
The bigger picture is hard to miss. Frontier labs have transformed from research organizations into infrastructure companies operating at a scale that rivals national budgets. The US government’s move to gain pre-release review rights over the most powerful AI systems signals that Washington sees this technology as too consequential to leave entirely to market forces.
