AI News Weekly Summary — February 23 – March 1, 2026 | Top Artificial Intelligence Updates This Week

1. Trump Bans Anthropic from US Government — Pentagon Issues “Supply Chain Risk” Designation

On February 27, Trump signed an executive order directing federal agencies to immediately cease using Anthropic technology. Defense Secretary Pete Hegseth designated the company a “Supply-Chain Risk to National Security” — a classification previously reserved for foreign adversaries.

The reason: Anthropic refused to lift two red lines — banning Claude from mass surveillance of Americans and from fully autonomous weapons. The Pentagon gave Anthropic a deadline of 5:01 PM ET Friday. The deadline passed without agreement.

Agencies have 6 months to phase out Anthropic products. Claude was the only AI model available on US military classified systems (contract: $200M, via Palantir).

Hours later, OpenAI announced a deal with the Pentagon to deploy its models on classified networks. Altman noted the agreement includes prohibitions on mass surveillance and human responsibility for use of force — “the same principles Anthropic was trying to protect.”

Anthropic announced it will legally challenge the supply chain risk designation.

This is unprecedented: the US government is punishing an AI company for the first time for refusing to drop safety guardrails. The paradox — OpenAI gets the contract with the same restrictions the Pentagon wouldn’t accept from Anthropic.

2. OpenAI Closes Record $110 Billion Round — $840 Billion Valuation

On February 27, OpenAI announced the largest private funding round in history: $110B. Investors: Amazon ($50B, with $15B upfront, $35B conditional on milestones), SoftBank ($30B), Nvidia ($30B).

Pre-money valuation: $730B, post-money: $840B. The round remains open — OpenAI expects additional investors.

Amazon also expanded its existing AWS agreement by $100B over 8 years. OpenAI separately signed a contract with Cerebras Systems for 750 MW of compute at over $10B for 3 years.

The two largest private AI funding rounds in history — Anthropic ($30B) and OpenAI ($110B) — closed in the same month. The scale of financing has moved beyond any historical precedent.

3. Meta Signs $100 Billion AMD Chip Deal — With 10% Equity Option

On February 24, Meta signed a deal with AMD to purchase MI450 chips worth up to $100 billion. The agreement covers 6 GW of data center capacity. First shipments (1 GW) begin in H2 2026.

Key: Meta gains an option for a 10% stake in AMD. This signals Meta is diversifying chip supply beyond Nvidia — and is willing to pay with a strategic equity position.

Meta enters the chip market not as a customer but as a potential co-owner. Nvidia faces serious competition in a hyperscaler’s portfolio for the first time.

4. Perplexity Computer — “Digital Worker” With 19 AI Models

On February 25, Perplexity launched Perplexity Computer — a platform orchestrating 19 specialized AI models (Claude Opus 4.6, GPT-5.2, Gemini, Grok, and more) as a “digital worker.”

How it works: it breaks tasks into sub-agents running in parallel — for hours or even months. Each agent has a browser, filesystem, and tool integrations. If one gets stuck, it spawns more agents to solve the problem.

Available to Perplexity Max subscribers. Pro expansion planned.

5. DeepSeek Preps V4 on Chinese Chips

DeepSeek is preparing to launch V4 — a multimodal model, its first major release since January 2025’s R1. The Hangzhou lab worked with Huawei and Cambricon to optimize V4 for Chinese chips, reducing dependence on Nvidia.

On February 11, DeepSeek quietly expanded its context window to 1M tokens and updated its knowledge cutoff. No official release date confirmed, but signals point to the coming weeks — just after the Two Sessions (March 4), underscoring DeepSeek’s status as a national AI champion.

DeepSeek V4 is a test of whether China’s chip ecosystem (Ascend + Cambricon) is ready for frontier models without Nvidia. If so — the embargo becomes an empty gesture.

6. Anthropic Launches Enterprise Plugins for Claude

Anthropic introduced industry-specific plugins allowing Claude to act directly within enterprise tools: Excel, PowerPoint, Google Drive, Gmail. Instead of returning instructions, Claude executes multi-step actions autonomously — generating reports, formatting presentations, sending emails.

Power users can design and train custom plugins tailored to specific business units.

7. Apple Siri with Gemini — Key Features Delayed

Apple planned to launch a new Siri powered by Gemini (1.2T parameters) on Google Private Cloud Compute with iOS 26.4 (March). However, some features — including full contextual awareness and cross-app integration — have been pushed to iOS 26.5 (May) or iOS 27 (September).

Tim Cook signals that Visual Intelligence will be the defining feature of AI wearables. Hardware launches expected from the week of March 2.

8. Google Relaunches Flow — AI Creative Studio

Google refreshed and expanded Flow — an AI creative studio integrating Whisk, ImageFX, Nano Banana 2, and Veo models. New features: lasso image editing, media collections, clip extension, camera movement controls.

Nano Banana 2 maintains consistency for up to 5 subjects across multiple images — critical for ad campaigns.

9. Micron Breaks Ground on $100 Billion Megafab

Micron Technology begins construction of a megafab in New York state — the largest private investment in the state’s history. Production: advanced DRAM and HBM critical for AI. Cost: $100B. ~50,000 jobs over 20+ years.

First facility construction: late 2026 – Q2 2028. Production from Q1 2029. Target: 40% of Micron’s total US DRAM output.

10. TSMC — AI Chip Revenue Growing at 60% CAGR Through 2029

TSMC forecasts AI chip revenue growth at 60% CAGR through 2029. Q3 2025 market share: 71% (vs. 64.9% a year earlier). 2026 capex: $52–56B — a record.

11. OpenAI Takes Over Pentagon Military AI Contract

Hours after the Anthropic ban, OpenAI announced a deal with DOD to deploy models on classified military networks. Altman confirmed the agreement includes bans on mass surveillance and requires human oversight for use of force.

The irony: OpenAI accepted the same restrictions the Pentagon refused from Anthropic. The difference — OpenAI didn’t try to enforce them as hard red lines.

12. Airbnb Deploys AI in Customer Support

Airbnb quietly rolled out an AI assistant in customer support across the US, Canada, and Mexico. The system already resolves about 1/3 of all tickets without human involvement.

13. Anthropic Study: AI Help Speeds Coding but Hurts Understanding

An Anthropic study found that while AI assistance speeds up coding, it significantly harms short-term conceptual understanding among developers learning new libraries. Developers with AI wrote faster but understood less of what they wrote.

14. Grok Correctly Predicted the Date of the Iran Military Operation

Grok (xAI) was the only AI model to correctly predict the launch date of the military operation against Iran — February 28, 2026. Other models missed.

15. OpenAI–Cerebras — 750 MW Contract Worth $10B+

OpenAI signed a deal with Cerebras Systems for 750 MW of compute over 3 years. Contract value: over $10B. Cerebras delivers wafer-scale chips — an alternative to Nvidia GPUs.

16. Google Responsible AI Report — AI as “Proactive Partner”

Google published its 2026 report, defining the transition from reactive tools to proactive partners capable of reasoning. Focus on agentic systems, scientific discovery, and safety.

17. Neuromorphic Computers Solve Physics Equations

Researchers demonstrated that neuromorphic computers modeled on the human brain can solve complex physics simulation equations — a task previously reserved for energy-hungry supercomputers.

18. Delhi Declaration — 70 Countries Sign AI Governance Principles

At least 70 countries signed the Delhi Declaration on AI — a document from the India AI Impact Summit defining principles for AI cooperation and governance. The first global AI document initiated by the Global South.

19. LinkedIn Redefines SEO for AI Search

LinkedIn reports -60% in non-brand B2B traffic from AI-powered search. The company abandoned traditional SEO metrics in favor of visibility-based measurements — tracking presence in AI-generated responses.

20. Prediction Markets — OpenAI Back in the Game

After the $110B round and Pentagon contract, prediction markets shift: OpenAI leads March coding specialization (86%). Google maintains long-term forecast leadership (June: 54%). Anthropic drops from 84% to 71% following the government ban.

Weekly Takeaway

The week of February 23 – March 1, 2026 is when AI geopolitics stopped being abstract. Trump banned Anthropic from government use for refusing to drop safety guardrails — and OpenAI took over the military contract hours later, with the same restrictions in the agreement. A paradox that defines the current stage of the AI race: rules exist, but enforcing them is a political question, not a technical one. On the financial front — $140B in new capital (OpenAI + xAI) in a single month, Meta enters AMD chips with a 10% equity option, and Micron builds a $100B megafab. DeepSeek V4 on Chinese chips will test whether the US embargo still means anything.

Sources: CNN, CNBC, NBC News, NPR, Axios, Fortune, TechCrunch, Bloomberg, Tom’s Hardware, The Motley Fool, Business Standard, PCWorld, 9to5Mac, Perplexity Blog, Anthropic Blog, OpenAI Blog, Google Blog