Anthropic and Claude: 2026 AI Powerhouse

In early 2026, the image of Anthropic as a cautious, safety-oriented "research lab" has effectively been replaced by its reality: a $380 billion enterprise software powerhouse.

Between a massive release of new models and a high-stakes standoff with the U.S. government, Anthropic is currently navigating the most volatile month in its history. Here is the breakdown of what is actually happening.

The Tech: Claude Now "Operates" Computers

Anthropic just released Claude 4.6 (Sonnet and Opus), and the focus has shifted from conversation to Computer Use.

While previous models could browse the web, Claude 4.6 is built to visually interpret a computer desktop. It doesn’t just call an API; it looks at a screen, moves the cursor, and clicks buttons.

To support this, Anthropic acquired Vercept this week, a startup that specialized in AI perception. The goal is to move Claude toward 100% reliability in navigating standard office software like spreadsheets, internal databases, and CRM tools that were not built for AI. Current benchmarks show Claude’s success rate in these environments has jumped from 15% to over 72% in the last year.

The Business: A $14 Billion Run-Rate

The company recently closed a $30 billion Series G funding round, valuing it at $380 billion. This puts Anthropic in the same valuation league as SpaceX and Coca-Cola.

The growth is driven by a pivot toward workhorse tools:

• Claude Code: This specialized engineering tool is now generating $2.5 billion in annual revenue, doubling its size since January 1st.
• Ad-Free Commitment: Anthropic made a point to announce that Claude will remain ad-free, positioning itself as a clean infrastructure provider for enterprises that don't want their data used to fuel a marketing engine.
• The 1M Token Context: Both new 4.6 models now support a one-million-token window, allowing businesses to feed entire technical libraries or legal archives into a single prompt with high retrieval accuracy.

The Conflict: The Pentagon Standoff

The most significant story right now isn't about code; it's about national security. Anthropic is currently the only AI provider allowed on the U.S. military’s classified networks, and tensions have reached a breaking point.

• The Venezuela Incident: Reports surfaced that Claude was utilized during the operation to capture Nicolás Maduro in January. This sparked a conflict between Anthropic’s red lines, specifically its ban on using AI for autonomous targeting or domestic surveillance and the Pentagon’s desire for "unfettered access."
• The Ultimatum: Defense Secretary Pete Hegseth has given Anthropic a deadline of Friday, 5:00 PM, to lift these restrictions. If they don't, the government has threatened to label Anthropic a "supply chain risk," which would effectively bar them from any federal contracts and potentially disrupt their private sector partnerships.

The Policy: Responsible Scaling 3.0

In the midst of this, Anthropic updated its Responsible Scaling Policy (RSP) to Version 3.0.

The new version is noticeably more pragmatic. It explicitly states that Anthropic will no longer delay development of powerful models unilaterally if it believes its competitors are moving forward. It’s a "competitor-contingent" policy: they will be as safe as the market allows them to be without losing their lead.

Critics have called this a safety loophole, but for Anthropic, it’s a necessary adjustment to a 2026 reality where AI is now central to global economic and military power.