AI News Roundup — March 5, 2026
OpenAI drops GPT-5.4, the Pentagon formally blacklists Anthropic as a supply-chain risk, and Oracle plans up to 30,000 layoffs to fund its AI pivot — a day dominated by escalating AI-military tensions and massive workforce restructuring.
OpenAI Releases GPT-5.4 with Pro & Thinking Variants
â OpenAI dropped GPT-5.4, its latest model update, featuring new Pro and Thinking model variants. The Thinking variant is designed for complex multi-step reasoning, while Pro targets enterprise use cases with improved reliability and tool use. The rapid cadence â just weeks after GPT-5.2 â signals OpenAI is accelerating its release schedule.
Pentagon Officially Labels Anthropic a "Supply-Chain Risk" â Anthropic Fights Back
â The Department of Defense formally designated Anthropic as a national security supply-chain risk, escalating weeks of tension after the company refused to remove safety guardrails for military use. Anthropic announced it will legally challenge the designation, calling it "unprecedented and politically motivated." This is the first time a major AI lab has been formally blacklisted by the US military.
Oracle Plans 20,000â30,000 Layoffs in AI Infrastructure Pivot
â Oracle is planning to cut up to 30,000 employees â one of the largest tech layoffs of 2026 â to fund its AI data center expansion. The company is carrying $100B+ in debt, partly from its OpenAI partnership commitments. Combined with Block's 4,000 cuts and Morgan Stanley's reductions, over 35,000 tech jobs have been cut in recent weeks.
AWS Launches AI Agent Platform for Healthcare
â Amazon Web Services introduced Amazon Connect Health, an AI-powered agent system for healthcare providers. It handles appointment scheduling, medical documentation, and patient identity verification â targeting the massive administrative burden in US healthcare.
Luma Unveils Creative AI Agents with "Unified Intelligence" Models
â Luma launched a new suite of creative AI agents powered by its "Unified Intelligence" models, designed for visual content creation, video editing, and creative workflows. The launch positions Luma as a competitor to Runway and Midjourney in the fast-growing creative AI space.
Cursor Rolls Out Agentic Coding Tools
â Cursor, the AI-powered code editor, launched new agentic coding capabilities that let its AI assistant autonomously plan and execute multi-file code changes, run tests, and iterate on failures. The release intensifies the AI coding wars with GitHub Copilot, Claude Code, and Windsurf.
US Considering Sweeping New Chip Export Controls
â The government is reportedly exploring expanded semiconductor export restrictions that could further limit China's access to advanced AI chips. The new controls would go beyond current limits on Nvidia and AMD hardware, potentially targeting lower-tier chips that Chinese firms have been using as workarounds. TechCrunch","published_at","2026-03-06T16:00:00.000Z","created_at","2026-03-06 16:01:22","actionData","errors"]