AI News Roundup: GPT-5.4 Lands With Computer Use, Anthropic Gains Cross-Industry Support, Apple’s AI Siri Nears Launch

GPT-5.4: OpenAI Ships Its Most Capable Model With Native Computer Use

OpenAI released GPT-5.4 on March 5, its most capable frontier model to date. The headline feature is native, state-of-the-art computer-use capabilities — agents can now operate desktops, click through applications, and execute multi-step workflows autonomously. The model supports a 1-million-token context window and ships in two variants: GPT-5.4 Thinking (for deep reasoning) and GPT-5.4 Pro.

On benchmarks, GPT-5.4 set a record 83% on OpenAI’s GDPVal test for economically valuable knowledge work and scored 75% on OSWorld-V for simulated desktop productivity tasks. OpenAI says individual claims in GPT-5.4 responses are 33% less likely to be false compared to GPT-5.2, and the model uses significantly fewer reasoning tokens to solve problems. It also incorporates full-fidelity vision supporting up to 10.24 million pixels per image.

Tech Industry Rallies Behind Anthropic in Pentagon Showdown

The standoff between Anthropic and the Pentagon intensified this week as the AI firm sought an emergency stay from the D.C. Circuit Court of Appeals. The Department of Defense designated Anthropic a “supply-chain risk” — a label historically reserved for foreign adversaries — after the company refused to let its AI be used for mass surveillance of Americans or autonomous weapons.

In a rare show of cross-industry solidarity, more than 30 employees from OpenAI and Google DeepMind signed a statement supporting Anthropic’s position, and Microsoft filed its own amicus brief. Anthropic’s CFO warned the designation could cost the company billions in 2026 revenue, with over 100 enterprise customers already raising concerns.

The rhetoric between AI CEOs has also sharpened: Anthropic CEO Dario Amodei called OpenAI’s approach “safety theater,” while Sam Altman fired back that it’s “bad for society” when companies abandon democratic norms. Meanwhile, Google has been quietly expanding its own Pentagon AI contracts, including a deal to provide AI agents to the DOD’s 3-million-person workforce.

Apple’s AI-Powered Siri Targets March Launch With Gemini Under the Hood

Apple is preparing to ship a completely reimagined version of Siri in iOS 26.4, powered by Google’s Gemini models in a deal reportedly worth $1 billion per year. The new Siri moves beyond simple command-and-response to become a context-aware AI assistant with on-screen awareness, meaning it can read what’s displayed on a user’s device and take action accordingly.

Demonstrated capabilities include extracting flight details from Mail, pulling reservation info from Messages, and making restaurant reservations from Safari without the user needing to copy-paste any information. Siri will also support multi-step task chaining and natural multi-turn conversations. However, some reports suggest certain features may slip to iOS 26.5 or later if Apple decides they aren’t ready.

OpenAI Acquires OpenClaw Creator as Agent Race Heats Up

OpenAI hired Peter Steinberger, the creator of OpenClaw, the open-source AI agent framework that became the most-starred project on GitHub in early March — surpassing both React and Linux with 68,000 stars. Steinberger will lead development of next-generation personal agents at OpenAI, while OpenClaw itself moves to an independent foundation with OpenAI as a financial sponsor.

OpenClaw distinguished itself by combining tool access, sandboxed code execution, persistent memory, and integration with messaging platforms like Telegram, WhatsApp, and Discord into a single package. The acquisition signals OpenAI’s aggressive bet that the future of AI is less about what models can say and more about what they can do. NVIDIA is also entering the space, announcing plans for an open-source enterprise agent platform called NemoClaw.

Google Drops Gemini 3.1 Flash-Lite: Speed and Cost Redefined

Google launched Gemini 3.1 Flash-Lite on March 3, the fastest and most cost-efficient model in the Gemini 3 family. Priced at just $0.25 per million input tokens and $1.50 per million output tokens, it delivers 2.5x faster time-to-first-token and 45% faster output speed compared to Gemini 2.5 Flash.

The model introduces adjustable “thinking levels” so developers can tune how much reasoning the system applies before generating a response. It’s available in preview on Google AI Studio and Vertex AI, targeting high-volume, cost-sensitive workloads like translation, content moderation, and UI generation.

Big Tech Commits $650 Billion to AI Infrastructure in 2026

Microsoft, Amazon, and Meta are projected to spend a combined $650 billion on AI infrastructure this year, a nearly 60% increase from 2025. Microsoft alone is on pace to spend more than $140 billion in fiscal year 2026, with CEO Satya Nadella pledging to increase AI capacity by over 80%.

Meta signed a multiyear, multigenerational partnership with NVIDIA to build hyperscale data centers deploying millions of Blackwell and Rubin GPUs. The spending surge comes as Morgan Stanley projects a net U.S. power shortfall of 9 to 18 gigawatts through 2028 — a 12% to 25% deficit — raising questions about whether power grids can keep up with AI’s appetite.

a16z’s Top 100 Gen AI Apps: ChatGPT Hits 900M Weekly Users

Andreessen Horowitz released its latest ranking of the top 100 generative AI consumer apps, and ChatGPT continues to dominate — now reaching 900 million weekly active users, up 500 million from a year ago. It remains 2.7x larger than the second-place Gemini on web traffic and 2.5x larger on mobile.

Paid subscriber growth tells a different story: Claude is growing paid subscribers at over 200% year-over-year, while Gemini leads at 258%. The report also expanded its definition to include products where generative AI has become core to the experience, adding names like CapCut, Canva, Notion, and Grammarly to the list.