The DOJ classified Anthropic as an "unacceptable national security risk" with a 40-page document, and on the same day, Axios reported that Anthropic's enterprise market share has surpassed OpenAI. The government penalizes it on principle, while the market rewards it for its products.

1 | The DOJ classifies Anthropic as an "unacceptable risk," but corporate clients are voting with their money for it

On Tuesday, the U.S. Department of Justice submitted a 40-page document to the U.S. District Court for the Northern District of California, stating that Anthropic has the ability to "disable or modify technology according to its own interests in wartime," and thus constitutes an "unacceptable risk" in the defense supply chain. The document used the term "Department of War," favored by the Trump administration, redefining Anthropic's ethical technology clauses as strategic vulnerabilities. Hearing on March 24.

On the same day, according to Axios, Anthropic's share of enterprise AI spending rose from less than 40% in early December to about 40% now, while OpenAI dropped from 50% to 27%. Anthropic has an annualized revenue of $19 billion, while OpenAI's is $25 billion, but according to Epoch AI’s calculations, Anthropic's annual growth rate is nearly three times that of OpenAI, and it is expected to surpass OpenAI mid-year on the current trajectory.

Morning reports covered court documents supporting Anthropic from the entire industry. The evening update was the government's formal response, along with market data that surfaced on the same day. The two "judgments" arrived simultaneously, with completely opposite directions. The government said, "AI companies with ethical guidelines are safety hazards," while corporate clients said, "AI companies with ethical guidelines are more reliable suppliers."

(Source: Bloomberg / Axios / Epoch AI / Al Jazeera)

2 | On the third day of NVIDIA GTC, Jensen Huang personally hosts the open-source camp's gathering

The last day of GTC focused not on hardware but on ecology. Jensen Huang personally hosted a roundtable on open models, sitting across from LangChain founder Harrison Chase, leaders from A16Z, AI2, Cursor, and Thinking Machines Lab. The core question discussed was whether open models can catch up with closed-source frontiers.

A more substantive action was the establishment of the Nemotron Coalition. This alliance brings together eight organizations: Cursor, LangChain, Mistral AI, Perplexity, Black Forest Labs, Reflection AI, Sarvam, and Thinking Machines Lab. The first project is to develop foundational models jointly by Mistral and NVIDIA to be trained on DGX Cloud and released to the open-source ecosystem.

Monday discussed hardware parameters, Tuesday focused on inference grids, and Wednesday on open-source alliances. Over the three days, NVIDIA is building not just a chip business, but a complete AI ecosystem based on its own computational resources. The underlying message of the open model alliance is that all participants must run on NVIDIA's computing power.

(Source: NVIDIA Newsroom / Tom's Hardware / TechCrunch)

3 | Mistral turns the table at GTC, Forge allows enterprises to train their own models from scratch

According to TechCrunch, Mistral launched the Forge platform at GTC, allowing enterprises to train exclusive models from scratch using Mistral's library of open-source weighted models. The first batch of clients includes Ericsson, the European Space Agency, and Singapore's DSO. Mistral Small 4 was also released, featuring a 119 billion parameter MoE architecture with only 6 billion active parameters, licensed under Apache 2.0, with 40% faster inference speeds than previous generations.

The logic of Forge is completely opposite to the API model of OpenAI and Anthropic. The latter sells black-box inference, while Forge sells complete control from data to weights. For sovereign-sensitive clients in Europe and Asia, the statement "ownership of the model is in my hands" is worth more than performance scores.

(Source: TechCrunch / VentureBeat / Dataconomy)

4 | AI layoffs transform from isolated incidents into industry templates

According to Fortune, analysts predict that Zuckerberg is "completing what Dorsey started." Jack Dorsey laid off 40% of Block's employees (4,000 people) at the end of February, forecasting that "most companies will make the same structural decisions within a year." Two weeks later, Atlassian laid off 1,600 people, over 900 of which were research and development positions. CEO Mike Cannon-Brookes stated, "It is dishonest to pretend that AI does not change the required skill sets and number of positions." Last week, Meta announced a plan to lay off 20% of its workforce, about 16,000 people.

Since 2026, tech industry layoffs have exceeded 55,000. Dorsey's template is the most radical version, cutting nearly half the staff, with stock prices rising 17%, signaling market rewards. As layoffs become synonymous with "AI transformation," replacing human costs with computational resources is no longer an isolated decision but a structural reset being replicated across the industry.

(Source: Fortune / TechCrunch / Bloomberg)

Also worth noting ↓

Bitrefill confirms a breach by North Korea's Lazarus group, with 18,500 purchase records leaked. According to CoinDesk, the attack originated from old credentials on an employee's laptop, allowing attackers to enter the production environment and steal funds from hot wallets. Bitrefill stated it will cover the losses using operational funds. The security vulnerabilities of crypto infrastructure companies are not on-chain, but on that employee's laptop that was not reclaimed in time. (Source: CoinDesk / Bitcoin Magazine)

Bitcoin ETFs have seen nearly $1 billion in inflows since March 9, with BTC consolidating around $74,000. The increase during this period was 12%, but on-chain data showed a surge in exchange inflows, with selling pressure near the $75,000 resistance level. Bitcoin Core released v31.0rc1 as a test version on Wednesday. (Source: Fortune / CoinGecko)

Funding in robotics exceeded $1.2 billion in a single week, with expectations to surpass $20 billion in 2026. The number of AI funding rounds exceeding $100 million generated in the first two and a half weeks of March has surpassed any similar period in venture capital history. Capital is overflowing from language models to automation in the physical world. (Source: AI Funding Tracker)

The FOMC interest rate decision will be announced tomorrow at 2 AM Beijing time. The market expects the rate to remain unchanged at 3.5%-3.75% (with over 92% probability), with a focus on the dot plot and Powell's press conference wording. Powell's term ends on May 15, and this will be his second-to-last press conference as Fed Chair. (Source: CNBC / Kiplinger)

Google Gemini's paid users grew 258% year-on-year, surpassing Claude's 200%. However, in terms of enterprise market share, Anthropic is widening the gap. The competition in AI is diverging into two different wars between consumer and enterprise markets. (Source: Axios / UCStrategies)

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。