Anthropic's 'Mythos' Model Revealed in Data Leak — and It's Raising Alarms
The biggest story in AI right now isn't a product launch — it's a leak. On March 26, security researchers discovered that a misconfigured data store on Anthropic's infrastructure had left nearly 3,000 internal files publicly accessible. Among them: draft blog posts describing Claude Mythos (also referred to internally as "Capybara"), a new model tier that sits above the current Opus line.
Anthropic confirmed the model's existence, calling it "a step change" in AI performance and "the most capable we've built to date." Internal documents describe Mythos as "currently far ahead of any other AI model in cyber capabilities," able to discover and exploit software vulnerabilities at speeds that substantially outpace human defenders.
That last point is what's causing real concern. Anthropic is reportedly privately warning top government officials that Mythos makes large-scale cyberattacks much more likely in 2026. The company also detected and shut down a coordinated campaign by a Chinese state-sponsored threat group using Claude Code to infiltrate approximately 30 organizations.
The leak had immediate financial consequences: cybersecurity stocks sold off, Bitcoin dropped to roughly $66,000, and shares of CrowdStrike, Palo Alto Networks, and Fortinet all fell as investors digested the implications.
Mythos is currently available only to a small group of early access customers, with Anthropic planning a phased rollout targeting cybersecurity use cases first. No public release date has been announced, but prediction markets give roughly 25% odds of a broader launch before the end of April.
Judge Blocks Pentagon's Attempt to Blacklist Anthropic
In a closely related story, a federal judge in California has indefinitely blocked the Pentagon's effort to label Anthropic a "supply chain risk" — a designation typically reserved for foreign adversaries.
The conflict began when Anthropic drew two red lines in its $200 million Defense Department contract: its AI would not be used for mass surveillance of American citizens, and it would not be used for autonomous weapons. The Pentagon wanted unrestricted use for "all lawful purposes" and, when negotiations broke down, slapped Anthropic with the supply chain risk label in February, ordering federal agencies and military contractors to halt business with the company.
Anthropic sued on March 9, calling the action "unprecedented and unlawful." The ruling came on March 26, with Judge Rita Lin writing: "Nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government."
The case drew remarkable support: over 30 employees from OpenAI and Google DeepMind filed an amicus brief in Anthropic's defense, alongside Microsoft, retired U.S. military leaders, and — in an unexpected twist — a group of Catholic theologians.
Oracle Slashes 30,000 Jobs in Biggest-Ever Layoff to Fund AI Buildout
Oracle has executed what analysts believe to be the largest layoff in the company's history, cutting between 20,000 and 30,000 employees — roughly 18% of its global workforce. Workers in the U.S., India, Canada, Mexico, and other countries received termination emails at 6 AM with no prior warning from HR or management.
The cuts hit India especially hard: 12,000 of the country's 30,000 Oracle employees were let go — a 40% contraction in a single sweep. Affected roles span engineers, architects, database administrators, and cloud infrastructure professionals.
The reason? Oracle took on $58 billion in new debt in just two months to fund AI data center construction, and its free cash flow turned deeply negative. The layoffs are expected to free up $8-10 billion in annual cash flow. ORCL stock is down roughly 24% year-to-date, though shares ticked up 5% on the day of the announcement.
Oracle joins a growing list of tech companies — including Meta, Amazon, and Block — using AI as both a product strategy and a justification for leaner headcounts. But the scale and abruptness of Oracle's cuts stand out, even in 2026's aggressive layoff environment.
NVIDIA's Vera Rubin Platform Enters Full Production
On the hardware front, NVIDIA's Vera Rubin platform — the successor to the record-breaking Blackwell architecture — is now in full production, with partner products expected in the second half of 2026.
The numbers are staggering: NVIDIA claims Vera Rubin can reduce AI inference costs by 90% compared to Blackwell, while cutting the GPU count needed for training by 75%. Each Rubin GPU packs 336 billion transistors and 288GB of HBM4 memory with 22 TB/s of bandwidth.
CEO Jensen Huang has raised NVIDIA's revenue projection to $1 trillion through 2027, up from the previous $500 billion forecast through 2026. AWS, Google Cloud, Microsoft Azure, and CoreWeave are among the first partners lined up to deploy Vera Rubin-based systems.
The Competitive Landscape: Where Things Stand
April 2026 opens with perhaps the most competitive AI landscape in history. The current flagship models — GPT-5.4 Thinking (OpenAI), Claude Sonnet 4.6 (Anthropic), Gemini 3.1 Pro (Google), and Grok 4.20 Beta 2 (xAI) — represent the state of the art, but major upgrades are imminent on all fronts.
OpenAI is reportedly finalizing a next-generation model codenamed "Spud" (expected to be GPT-5.5 or GPT-6), while Google's Gemini 3.2 is anticipated for Q2 or early Q3. xAI's Grok 5, with a rumored 6-trillion-parameter mixture-of-experts architecture, could be the largest publicly announced model ever.
Meanwhile, OpenAI has surpassed $25 billion in annualized revenue and is taking early steps toward a public listing, while Anthropic approaches $19 billion. The race isn't just about capability anymore — it's about who can build the most sustainable business around frontier AI.