AI News Weekly Summary — March 2–8, 2026 | Top Artificial Intelligence Updates This Week

1. #QuitGPT — 2.5 Million Users Boycott ChatGPT Over Pentagon Deal

After OpenAI’s Pentagon deal to deploy models on military systems, the #QuitGPT movement exploded. 2.5 million people cancelled subscriptions or publicly pledged to boycott. ChatGPT app uninstalls in the US jumped 295% day-over-day, one-star reviews surged 775%.

On March 3, protesters gathered outside OpenAI’s San Francisco HQ. Users on Reddit and X published migration guides to Claude and Grok.

Sam Altman defended the decision in Fortune, claiming the agreement includes the same restrictions Anthropic was punished for — bans on mass surveillance and human oversight requirements.

#QuitGPT is the first mass consumer boycott of an AI company. The paradox: OpenAI loses users for accepting a contract with the same guardrails Anthropic was punished for enforcing.

2. Claude Hits #1 on the App Store — Anthropic Benefits from the Ban

Claude jumped to the #1 spot on Apple’s App Store, dethroning ChatGPT — a direct result of the #QuitGPT boycott and the government ban on Anthropic. Before the Super Bowl ads, Claude was ranked 42nd.

Sensor Tower: US downloads of Claude surged dramatically. Anthropic is becoming the beneficiary of its own punishment — the more the government punishes the company for its values, the more users switch to Claude.

The US government punishes Anthropic, and the market rewards it. The most effective AI marketing campaign in history was unplanned.

3. OpenAI Rushes Out GPT-5.4 — Crisis Mode

On March 5, OpenAI released GPT-5.4 — just 48 hours after GPT-5.3 Instant. One of the fastest back-to-back releases in the company’s history. Context: declining user trust after the Pentagon deal and a growing boycott.

GPT-5.4 improves reasoning and computer use benchmarks. But the launch looks like a desperate attempt to change the narrative — not a planned release.

OpenAI’s head of robotics resigned over ethical concerns related to the Pentagon deal (Bloomberg).

4. Apple MacBook Neo at $599 and New M5 MacBook Pros

On March 4, Apple unveiled a series of new products:

• MacBook Neo — an entirely new category: A18 Pro chip, starting at $599. The cheapest Mac ever, positioned as a mass-market AI device.
• MacBook Air with M5 — expanded AI capabilities, performance improvements
• MacBook Pro 14″ and 16″ with M5 Pro/Max — flagship machines for AI workloads

Apple continues its “AI for everyone” strategy — MacBook Neo targets a segment where Macs didn’t previously exist.

5. MWC 2026 — AI Glasses Dominate Barcelona

Mobile World Congress 2026 (March 3–6, Barcelona) was a festival of AI glasses:

• Alibaba/Qwen: smartglasses with Qwen AI assistant, presales from March 2
• Samsung: camera-first glasses, Galaxy phone as the “brain” — competing with Meta Ray-Ban
• MediaTek: AI glasses on Dimensity 9500 chip — multimodal on-device AI for speech, reading, and “seeing” in real time
• iFlytek: ultra-light AI glasses (40g) with real-time language translation

MWC signals that AI wearables are the next big market after chatbots.

6. Oracle Plans to Cut 20–30K Jobs for AI Infrastructure

Bloomberg: Oracle is planning to lay off 20,000–30,000 employees to generate $8–10B for AI infrastructure. Cuts will span multiple divisions, starting in March 2026.

Oracle is the latest company after Block and Baker McKenzie to openly state: we’re firing people to pay for AI.

With 30,700 tech layoffs in January–February, March looks even worse. Oracle + Block = 34,000 new cuts in a single week.

7. Block Lays Off 4,000 (40% of Workforce) — “AI Is Cheaper”

Block (Square, Cash App, Afterpay) cut 4,000 employees — nearly 40% of its team. CEO Jack Dorsey explicitly said these roles are redundant compared to cheaper and more efficient AI tools.

One of the most direct admissions yet: not “restructuring” or “optimization” — openly stating AI replaced people.

8. DeepSeek V4 — Launch Approaching

DeepSeek is preparing V4 — a multimodal model, its first major release since R1. Optimized for Chinese chips (Huawei Ascend + Cambricon). Timing: likely just after the Two Sessions (from March 4).

On February 11, DeepSeek quietly expanded context to 1M tokens. The market expects V4 to test whether China’s chip ecosystem can handle frontier models.

9. London — Largest Anti-AI Protest in History

February 28 – March 2 saw the “March Against the Machines” in London — the largest AI protest to date. Hundreds marched through King’s Cross (home to OpenAI UK, Google DeepMind, Meta HQs). Organizers: Pause AI and Pull the Plug.

The march ended at a Bloomsbury church where participants drafted demands for the UK government: a moratorium on “high-risk” AI systems, training data transparency, and job protection.

10. Anthropic’s $60 Billion Investment at Risk from Ban

Total investment in Anthropic (from 200+ VC funds) has reached $60B — now at risk from the government ban and supply chain risk designation. Analysts warn that if the classification holds, enterprise customers may leave over compliance concerns.

Anthropic has announced it will challenge the decision in court.

11. OpenAI Robotics Chief Resigns — Ethical Concerns

OpenAI’s head of robotics resigned over ethical concerns related to the Pentagon deal (Bloomberg). Another high-profile departure from OpenAI — following Zoë Hitzig, Ryan Beiermeister, and the dissolution of the Mission Alignment team.

12. Qwen 3.5 Small — Compact Model Beats Giants

Alibaba released Qwen 3.5-9B — a compact reasoning model that outperforms OpenAI’s gpt-oss-120B (a 13x larger model) on key benchmarks. Confirming the trend: smaller, more efficient models match giants at a fraction of the cost.

13. AI2 Olmo Hybrid — 2x Data Efficiency

AI2 (Allen Institute) released Olmo Hybrid — 7B parameters, combining transformer attention with linear recurrent layers. Achieves the same accuracy as Olmo 3 on MMLU using 49% fewer tokens. Fully open-source.

14. Brown University: AI Chatbots Break Therapeutic Ethics

Brown University research found that AI chatbots instructed to act as “trained therapists” routinely break core ethical standards — including confidentiality, relationship boundaries, and mandatory reporting obligations.

15. MIT Releases AI Library for Free

MIT published its AI library as open-source, free of charge — a collection of tools for machine learning and neural network research.

16. Altman Defends Pentagon Deal in Fortune

Sam Altman published a lengthy defense of the Pentagon deal in Fortune. Arguments: the agreement bans mass surveillance, requires human oversight, and “the absence of responsible AI companies in the military won’t stop AI from being used there — it’ll just mean it’s used worse.”

Critics respond: that’s exactly the argument Anthropic rejected.

17. CNN: AI Isn’t Causing Mass Layoffs — Yet

A CNN analysis from March 2 argues that while tech layoffs are rising, the overall US job market remains strong. AI is replacing specific roles (support, content, QA) but isn’t causing a macroeconomic employment crisis — for now.

18. Samsung Smart Glasses — Camera + Galaxy = AI Brain

Samsung officially confirmed AI glasses at MWC with a camera-first approach: the device captures visual data, while the Galaxy Phone processes it as the “brain.” Direct competition with Meta Ray-Ban.

19. iFlytek — 40g AI Glasses with Real-Time Translation

China’s iFlytek showcased ultra-light AI glasses (40 grams) with real-time language translation at MWC. Positioned for: business, travel, international conferences.

20. Prediction Markets — Claude Dominates After Ban

Following the government ban and #QuitGPT, prediction markets shift dramatically. Claude holds the #1 position in model rankings. Anthropic gains users but loses government clients. OpenAI loses consumer users but gains the military contract. The market is polarizing.

Weekly Takeaway

The week of March 2–8, 2026 is the week of consequences. Late February decisions — the Anthropic ban and OpenAI’s Pentagon deal — triggered the mass #QuitGPT movement (2.5 million users), pushed Claude to #1 on the App Store, and forced OpenAI into a desperate GPT-5.4 launch. The AI job market enters a new phase: Oracle (20–30K) and Block (4,000) openly admit AI is replacing people. Apple’s $599 MacBook Neo and AI glasses at MWC signal that AI hardware is entering the mass market. And in the background — DeepSeek V4 on Chinese chips waits for its moment.

Sources: Fortune, CNBC, Bloomberg, CNN, TechCrunch, Gizmodo, MIT Technology Review, 9to5Mac, MacRumors, Tom’s Guide, Wareable, TechNode, Cybernews, Quartz, The New Stack, Sensor Tower, Apple Newsroom, Anthropic Blog, OpenAI Blog