Latest AI News Cycle for Thursday, February 12, 2026

• Microsoft pursues AI self-sufficiency to reduce reliance on OpenAI: Microsoft's Mustafa Suleyman announced a strategic shift toward building in-house enterprise and healthcare models. This represents a major decoupling from their primary partner, OpenAI, potentially reshaping the competitive landscape of foundational models.
• Zhipu AI launches GLM-5 open-weights SOTA model: The release of GLM-5 marks a significant technical breakthrough for Chinese AI, achieving state-of-the-art performance in open-source benchmarks. The launch caused a 30% surge in related Chinese technology stocks and intensifies the global open-weights race.
• Pentagon demands AI access on secure classified networks: The U.S. Department of Defense is pressuring OpenAI and Anthropic to deploy their tools on air-gapped, classified networks. This move signals a massive shift in military adoption of LLMs for sensitive intelligence and tactical operations.
• Meta faces audit flags over $27B Hyperion project financing: Ernst & Young raised red flags regarding Meta's use of off-balance-sheet financial engineering for its $27 billion Hyperion data center. This highlights the immense financial risks and accounting complexities involved in the current AI infrastructure arms race.
• Elon Musk reorganizes xAI ahead of potential blockbuster IPO: Following a merger with SpaceX entities, Musk is restructuring xAI to prepare for a public offering. This reorganization consolidates his AI interests and positions xAI as a major capital-intensive competitor to OpenAI and Anthropic.
• Coinbase launches Agentic Wallets for autonomous AI crypto transactions: Coinbase has introduced infrastructure that allows AI agents to manage crypto wallets and conduct transactions without human intervention. This bridges the gap between agentic reasoning and real-world economic utility in the Web3 space.
• Amazon developers reject internal Kiro tool in favor of Claude: Internal reports indicate over 1,500 Amazon staff are pushing back against the company's proprietary Kiro coding assistant, preferring Anthropic's Claude Code. This highlights the growing influence of developer preference in enterprise tool adoption and skill requirements.
• Anthropic funds $20M Super PAC to oppose OpenAI's regulation: Anthropic is donating $20 million to a group pushing for strict AI transparency and guardrails. The move creates a direct political confrontation with OpenAI-backed PACs, signaling that regulatory capture is now a multi-million dollar battlefield.