AI News Weekly Summary — March 9–15, 2026 | Top Artificial Intelligence Updates This Week

1. Anthropic Sues Pentagon — First Amendment Lawsuit Over Supply Chain Risk Label

On March 9, Anthropic filed two federal lawsuits against the Trump administration — in California and Washington, D.C. The company alleges the Pentagon illegally retaliated after Anthropic refused to remove guardrails banning mass surveillance of US citizens and autonomous weapons.

The Pentagon wanted to use Claude for “all lawful purposes” — Anthropic drew two red lines and was punished for it. Under Secretary of Defense Emil Michael told Bloomberg: “The talks are over. We’re moving on.”

Court filings show over 100 enterprise customers contacted Anthropic about the designation. The company estimates risk of losing hundreds of millions to billions of dollars in 2026 revenue.

Anthropic is choosing the courtroom over compromise. This is the first time an AI company has sued the US government for punishing it over safety standards — a precedent that could define the boundary between private sector and Pentagon.

2. 30+ OpenAI and Google Employees Defend Anthropic in Court

More than 30 employees from OpenAI and Google DeepMind, including Google chief scientist Jeff Dean, filed an amicus brief supporting Anthropic. They warn that blacklisting a leading AI company “will undoubtedly have consequences for the United States’ industrial and scientific competitiveness.”

Quote from the filing: “This effort to punish one of the leading U.S. AI companies will undoubtedly have consequences for the United States’ industrial and scientific competitiveness in the field of artificial intelligence and beyond.”

Industry rivals standing together against the government — unprecedented. Even at OpenAI, which signed the Pentagon contract, employees publicly defend the competition.

3. Google Deploys Gemini AI Agents for 3 Million Pentagon Workers

One day after Anthropic’s lawsuit, Google announced it will provide Gemini-based AI agents to the Pentagon’s entire 3-million-person workforce — initially for unclassified work.

8 pre-built agents automate tasks: meeting summaries, budget building, checking actions against defense strategy. Staff can also create custom agents using natural language.

Google’s chatbot is already used by 1.2 million DoD employees, who have run 40 million prompts. Discussions are underway to expand to classified and top-secret networks.

4. Microsoft Copilot Cowork — Agents Powered by Anthropic’s Claude

On March 9, Microsoft announced Copilot Cowork — a new layer of M365 Copilot for multi-step, long-running tasks. Key detail: the agentic engine is powered by Anthropic’s Claude models, not OpenAI.

Copilot Cowork plans tasks and executes them across Outlook, Teams, Excel — resolving calendar conflicts, compiling briefings, building launch plans with competitive analysis. Available in Research Preview, broader launch in late March.

Microsoft pays billions to OpenAI, but chose Claude for its flagship agentic product. That says more about the market than any benchmark.

5. Perplexity Computer Goes Enterprise — 20 Models, Slack, 400+ Integrations

At its Ask 2026 conference, Perplexity announced Computer for Enterprise — an AI agent combining 20 specialized models with integrations for Snowflake, Salesforce, HubSpot, and 400+ other platforms.

Employees can query @computer directly in Slack channels. The company claims its enterprise agent completed 3.25 years of work in 4 weeks during testing.

Additionally, Perplexity unveiled Personal Computer — software running on Mac mini that gives the agent persistent access to local files, apps, and sessions.

6. NVIDIA and Thinking Machines Lab — Gigawatt Compute Deal

On March 10, NVIDIA announced a strategic partnership with Thinking Machines Lab (founded by former OpenAI CTO Mira Murati). The deal includes deploying at least 1 gigawatt of Vera Rubin systems starting in 2027.

NVIDIA also made a “significant investment” in the company (amount undisclosed). Thinking Machines has raised over $2 billion from Andreessen Horowitz, Accel, NVIDIA, and AMD Ventures.

A gigawatt of compute is a threshold only the largest labs approach. Murati isn’t building tools on others’ models — she’s building a frontier competitor to OpenAI.

7. NVIDIA GTC 2026 — What’s Coming March 16–19

NVIDIA’s biggest conference starts March 16 in San Jose. Jensen Huang delivers the keynote Monday at 11 AM PT (free stream).

Expected announcements:

• Details on Vera Rubin architecture (Blackwell successor)
• N1 and N1X chips for Windows laptops
• NemoClaw agentic platform for enterprise AI agent deployment
• New CPUs optimized for agentic AI inference

CNBC: CPU takes center stage over GPU — NVIDIA bets on inference processors for AI agents.

8. Atlassian Cuts 1,600 Jobs (10%) — “AI Changes the Skill Mix We Need”

On March 11, Atlassian announced layoffs of 1,600 employees (~10% of workforce) to “self-fund” investments in AI and enterprise sales. The cuts will cost the company $225–236M.

CEO Mike Cannon-Brookes: “Our approach is not ‘AI replaces people.’ But it would be disingenuous to pretend AI doesn’t change the mix of skills we need or the number of roles required in certain areas.”

CTO Rajeev Rajan steps down on March 31.

After Block (4,000), Oracle (20–30K), and Morgan Stanley (2,500) — Atlassian joins the wave. Over 55K tech layoffs in 2026, with 9,200+ explicitly attributed to AI.

9. Morgan Stanley: AI Breakthrough in 2026 — World Isn’t Ready

On March 13, Morgan Stanley published its “Intelligence Factory” report warning: a transformative AI leap will happen in H1 2026, and most of the world isn’t prepared.

Key projections:

• US power shortfall: 9–18 gigawatts through 2028 (12–25% deficit)
• Companies converting Bitcoin mines to HPC centers
• “15-15-15” model: 15-year leases, 15% yields, $15/watt net value
• AI as a deflationary force — replicating human work at a fraction of the cost

10. HIMSS 2026 — AI Agents Enter Hospitals at Scale

At HIMSS in Las Vegas (March 9–13), major tech companies showcased healthcare AI agents:

• Epic: 85% of customers using Epic AI. New Agent Factory lets hospitals build custom agents. Three main agents: Art (clinical notes), Penny (billing), Emmie (patient questions)
• Oracle Health: clinical note generation agent across 30 specialties, covering ER and inpatient
• Amazon Health AI: triage agent on Bedrock, connected to Health Information Exchange (patient history)
• Google and Microsoft: custom AI personas for healthcare

11. JPMorgan Raises Tech Budget to $19.8B — $1.2B for AI

JPMorgan Chase increased its 2026 technology budget to $19.8B (+10% YoY), with $1.2B dedicated to AI. Over 400 AI use cases are already running across the firm.

CEO Jamie Dimon calls AI an “existential priority.” Investments cover cloud infrastructure, cybersecurity, risk analysis, and fraud detection.

12. Trump vs. State AI Laws — March 11 Deadlines

Trump’s executive order set March 11 as the deadline for two key actions:

• FTC: issue a policy statement classifying state-mandated bias mitigation as “per se deceptive trade practice”
• Secretary of Commerce: publish a review of state AI laws identifying “overly burdensome” regulations

DOJ established an AI Litigation Task Force to challenge state AI regulations in federal court. States with “restrictive” AI laws may lose BEAD funding.

The Trump administration treats AI regulation like an arms race — whoever restricts AI, loses. Will the FTC actually call bias mitigation “deceptive”? That would set a far-reaching precedent.

13. DeepSeek V4 — Still “Just Around the Corner”

DeepSeek V4 remains the most anticipated open-source model of 2026. After multiple delays (February, late February, early March), on March 9 Chinese media reported a website update users called “V4 Lite” — but DeepSeek hasn’t officially confirmed anything.

Expected specs: trillion-parameter MoE, native multimodality, 1M+ token context, optimized for Chinese chips (Huawei Ascend, Cambricon). Claimed benchmark results have not been independently verified.

14. AI2 MolmoBot — Robot Trained Entirely in Simulation

On March 12, AI2 (Allen Institute) released MolmoBot — an open robotic manipulation model trained entirely on simulation data (1.8M trajectories). Achieves zero-shot transfer — works in the real world without fine-tuning.

Two platforms: Rainbow Robotics RB-Y1 (mobile manipulator) and Franka FR3 (tabletop arm). Full stack open-source: data, generation pipelines, training code.

Also released: MolmoSpaces — a simulation ecosystem for embodied AI research.

15. Siri 2.0 — Coming, But Slowly

Apple targets the revamped Siri for launch with iOS 26.4 (March 2026), but some features are pushed to May (iOS 26.5) or even iOS 27 (September). Siri’s architecture is being replaced with an LLM-based system.

Key detail: the engine is powered by Google Gemini — in a multi-year deal worth ~$1B/year. Internal testing revealed issues with the hybrid architecture (old + new system) — failures in ~1/3 of test cases.

16. 78 State AI Chatbot Bills Active Across 27 States

Just 6 weeks into the 2026 legislative session, 78 AI chatbot bills are active in 27 states. Topics: child safety, election deepfakes, AI content labeling, developer liability.

Virginia passed SB141 — requiring AI-generated political content to carry a disclaimer. Vermont signed a law on synthetic media in election campaigns.

17. 2026 Tech Layoffs: Over 55,000 and Counting

Total 2026 tech layoffs have surpassed 55,775 people (166 companies, averaging 754 per day). Over 9,200 explicitly attributed to AI and automation.

March: Atlassian (1,600), Morgan Stanley (2,500), Block (4,000), Oracle (20–30K planned). Amazon, Pinterest, and others openly point to AI as the reason for cuts.

CBS News: “More companies are pointing to AI as they lay off employees” — no longer a trend, it’s the standard.

18. Cognizant: “Plug-and-Play AI Is a Myth”

On March 10, Cognizant published research stating: enterprise AI deployment is not plug-and-play. Most organizations need 12–18 months to integrate AI into core processes, with ROI visible only after 2–3 years.

The report warns against “AI washing” — companies announcing AI transformation while using it only for simple chatbots, not actual process change.

19. 60% of Consumers Use AI Weekly — But Only 13% Trust It

New data: 6 in 10 consumers use AI tools weekly, but only 13% report full trust. AI-generated images produce 70% more content at the same cost, slashing ad spend.

20. Up Next: Will NVIDIA GTC Change the Game?

March 16–19 in San Jose will be the AI world’s epicenter. GTC 2026 is where NVIDIA traditionally sets the industry’s direction. Vera Rubin, laptop chips, NemoClaw agentic platform — Jensen Huang has a lot to show.

NVIDIA stock: Jim Cramer (CNBC) says the company is “in the sweet spot” — but Motley Fool asks whether GTC will be enough to boost the stock after its recent pullback.

---

Weekly Takeaway

The week of March 9–15, 2026 is the week of agents and lawsuits. Anthropic sues the Pentagon, and rivals — OpenAI and Google — stand by its side. Meanwhile, Google deploys Gemini agents to 3M Pentagon workers, Microsoft builds Copilot Cowork on Claude, and Perplexity ships Computer to enterprise. HIMSS shows AI agents entering hospitals at scale. On the jobs front — another wave: Atlassian, Morgan Stanley, 55K+ total in 2026. And just around the corner — NVIDIA GTC, which could change the rules of the game.

---

Sources: Fortune, TechCrunch, Bloomberg, CNBC, CNN, Axios, NPR, CBS News, VentureBeat, The Hill, Gizmodo, NVIDIA Blog, Microsoft Blog, Perplexity, Allen Institute, STAT News, Fierce Healthcare, American Banker, Transparency Coalition, GeekWire