AI News Roundup: Anthropic Accuses Chinese Labs of Industrial-Scale Model Theft as Pentagon Demands AI Guardrails Removed
Anthropic dominated the news cycle today from every angle: accusing three Chinese AI labs of industrial-scale model theft, launching a production-ready enterprise agents platform that sent software stocks surging, facing a Pentagon ultimatum to strip safety guardrails from Claude, and kicking off a $6 billion employee share sale at a $350 billion valuation. Meanwhile, Meta signed the largest AI chip deal in history and the Fed warned that AI-driven unemployment may be beyond its ability to fix.
Here’s everything that matters from February 24, 2026.
The Big Story: Anthropic Says DeepSeek, MiniMax, and Moonshot Stole Claude’s Capabilities at Industrial Scale
Anthropic published a detailed report revealing that three Chinese AI labs — DeepSeek, Moonshot, and MiniMax — ran coordinated campaigns to extract Claude’s capabilities through model distillation. The numbers are staggering: approximately 24,000 fraudulent accounts generated over 16 million exchanges with Claude, with MiniMax alone responsible for more than 13 million of those interactions.
The most provocative detail: Anthropic observed tasks where Claude was used to generate censorship-safe alternatives to politically sensitive queries, suggesting the distilled outputs were being used to train DeepSeek’s models to navigate China’s content restrictions. Anthropic warned that the resulting distilled models may lack the safety guardrails built into Claude, creating potential national security risks.
This is the most concrete accusation yet from a frontier AI lab about systematic model theft, and it raises uncomfortable questions about the enforceability of terms of service when the economic incentives for distillation are this strong. Fortune, CNBC, and Semafor all covered the story extensively.
Today’s Top Stories
Pentagon Gives Anthropic Until Friday to Drop Claude’s Safety Guardrails
Defense Secretary Pete Hegseth met with Anthropic CEO Dario Amodei at the Pentagon, delivering an ultimatum: grant the military unfettered access to Claude by Friday or face contract termination and designation as a supply chain risk. The dispute centers on Anthropic’s refusal to remove guardrails around AI-controlled weapons systems and mass domestic surveillance. Claude is currently the only AI model available in the military’s classified systems — though that may change, as xAI signed a deal to put Grok into classified systems under an “all lawful use” standard.
Meta Signs $100B+ AMD Deal in Pursuit of “Personal Superintelligence”
Meta announced a multi-year agreement with AMD to power its AI infrastructure with up to 6GW of AMD Instinct GPUs (MI540 series), with shipments beginning in the second half of 2026. The deal could exceed $100 billion in total value, and AMD received a performance-based warrant for up to 160 million Meta shares. AMD stock jumped 9% on the news. Zuckerberg called it “an important step” toward “personal superintelligence” — a phrase that now appears to be Meta’s official framing for its AI ambitions.
Anthropic’s Enterprise Agents Launch Sends Software Stocks Surging
Anthropic held “The Briefing: Enterprise Agents”, transitioning Claude Cowork from research preview to production-ready enterprise offering with plugins for Google Drive, Gmail, DocuSign, FactSet, S&P Global, and Slack. The market reaction was dramatic: Thomson Reuters surged as much as 14% — its biggest intraday gain in 26 years — after its CoCounsel AI legal tool was highlighted. The S&P 500 advanced 0.8%. On the flip side, IBM fell over 10% as Claude’s COBOL modernization tools threatened a key consulting revenue stream.
MatX Raises $500M to Build an AI Chip Rival to Nvidia
AI chip startup MatX, founded by two Google semiconductor alumni, raised more than $500 million in a round led by Jane Street and Leopold Aschenbrenner’s Situational Awareness fund, with participation from Marvell Technology and the Stripe co-founders. The company plans to complete its chip design this year with TSMC manufacturing and shipments beginning in 2027. It’s a big bet that there’s room for a third player beyond Nvidia and AMD in the AI silicon market.
Fed Governor Warns AI Could Cause Unemployment Beyond the Fed’s Reach
Federal Reserve Governor Lisa Cook warned that AI could cause structural unemployment that the Fed’s standard monetary policy tools cannot fix. She pointed to declining demand for coding jobs as evidence that the AI labor transition has already begun. Separately, Dallas Fed research showed the dual nature of AI’s labor impact: wages in computer systems design rose 16.7% since fall 2022, but employment in AI-exposed sectors is trailing the broader economy.
Quick Hits
- Big Tech Spending: Alphabet, Amazon, Meta, and Microsoft are collectively on track to invest $650 billion in AI infrastructure in 2026, up from $410 billion in 2025. Bridgewater warns the boom is entering “a more dangerous phase.”
- Anthropic Valuation: Anthropic kicked off a $5-6 billion employee share sale at a ~$350 billion valuation, with a potential IPO this year.
- OpenAI Enterprise Push: OpenAI launched “Frontier Alliances” — partnerships with McKinsey, BCG, Accenture, and Capgemini to accelerate corporate AI adoption.
- Interpretable AI: GuideLabs released Steerling-8B, an 8B-parameter model that traces every token back to its source data and 33,000+ supervised concepts, performing near Llama-8B on 10x less training data.
- Wolfram CAG: Stephen Wolfram announced “Computation-Augmented Generation” — integrating Wolfram Language into LLM workflows via MCP servers for dynamic computational content instead of static document retrieval.
- AI Chip Funding: Dutch startup Axelera AI raised $250 million backed by BlackRock for edge AI inference chips — the largest AI semiconductor investment in EU history.
- Firefox AI Kill Switch: Mozilla released Firefox 148 with a master toggle that disables all AI features in one click (438 points on HN).
- Goldman on AI GDP Impact: Goldman Sachs reported that AI contributed “basically zero” to US economic growth in 2025 despite hundreds of billions in capex.
- Google for Teachers: Google partnered with ISTE+ASCD to deliver free AI literacy training to all 6 million U.S. educators.
- F-35 AI: Lockheed Martin completed the first in-flight test of AI-enhanced combat identification on the F-35, retraining models between flights in minutes.
Anthropic at the Center of Every AI Debate
Today’s news painted a remarkably complete picture of the tensions defining the AI industry right now. Anthropic is simultaneously building the enterprise agent platform the market wants, confronting state-sponsored model theft, resisting government pressure to weaponize its technology, and preparing for an IPO at a valuation that would make it one of the most valuable private companies in history. The fact that one company sits at the intersection of all four of those stories in a single day tells you something about how concentrated the frontier AI ecosystem has become.
The Meta-AMD deal and the $650 billion collective Big Tech spend signal that the infrastructure buildout is accelerating, even as Goldman Sachs says AI has contributed essentially nothing to GDP growth so far. That gap between investment and measurable economic return is the story to watch in 2026 — and the Fed is already warning it may not have the tools to manage what happens if the transition gets bumpy.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
Get Carly Today →Or try our Free Group Scheduling Tool
