Anthropic’s Safety–Commerce Contradiction: The $965B AI Giant That Wants Everyone to Slow Down

These two realities — the safety evangelist and the hyper-aggressive commercial player — are now inhabiting the same company, and the friction is becoming impossible to ignore.

The pause call: what Anthropic actually said

Anthropic’s June 4 post was authored jointly by the head of its internal research institute and its head of policy . It disclosed internal metrics showing that the capabilities of Anthropic’s most advanced models are improving rapidly, and it warned that these advances appear to be on a path toward recursive self-improvement — a threshold some AI insiders view as a potential marker of danger and enormous societal upheaval .

The company did not propose a specific regulatory mechanism or treaty. Instead, it called for a globally coordinated agreement among AI labs to temporarily pause, or at least slow down, the pace of frontier model development .

Some market participants immediately interpreted the call as a marketing strategy or an attempt at regulatory capture — a company that has already achieved a lead asking everyone else to freeze in place .

The IPO and the $965 billion valuation

The financial timeline makes the contradiction vivid:

• May 28, 2026: Anthropic closes a $65 billion Series H round at a $965 billion valuation, led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital . It discloses a $47 billion annualized revenue run rate .
• June 1, 2026: Anthropic confidentially files a draft S-1 registration statement with the SEC, formally beginning the IPO process . The company cites high computing costs as the primary driver behind going public . Analysts expect a public debut in the second half of 2026, potentially in the fall .
• June 4, 2026: Anthropic publishes its global pause call .

Calling for an industry-wide slowdown while simultaneously raising record sums to build more compute, sell more models, and go public is the central tension. The company needs enormous capital to maintain its trajectory — and it’s simultaneously arguing that the trajectory itself may need to be capped.

Withholding Mythos: a genuine safety sacrifice

Earlier in 2026, Anthropic made a decision that appeared to put safety ahead of commercial interest. On April 7, 2026, it announced it would not release Claude Mythos — its most capable model — to the public after internal testing showed it could find high-severity vulnerabilities in major operating systems and browsers .

The model had identified thousands of previously unknown zero-day vulnerabilities, including a 27-year-old bug in OpenBSD and a 16-year-old flaw in FFmpeg that automated testing tools had failed to detect despite running the affected code line five million times . Anthropic restricted access to approximately 11 partners under a limited-access program called "Project Glasswing" .

This was unprecedented among major AI labs and represented a significant commercial sacrifice: Anthropic was voluntarily leaving revenue on the table by not shipping its best model to paying customers, at a time when rivals were releasing their own frontier models widely .

Dropping the 2023 safety pledge

In February 2026, however, Anthropic quietly abandoned the commitment that had been the cornerstone of its brand as the "safe" AI lab. Its Responsible Scaling Policy had included a pledge to pause development if its own AI became too dangerous . The company removed that commitment entirely .

The activist group Stop the AI Race documented the shift, noting that Holden Karnofsky, former CEO of Open Philanthropy and now at Anthropic, acknowledged that there was "enormous pressure" to downplay risks because triggering a pause would hurt the company . The solution was to remove the commitment .

That sequence — dropping a core safety pledge just months before filing for a historic IPO — strongly suggests that when the tension between safety rhetoric and commercial pressure became concrete, the commercial imperative won.

What the tension actually means

The four moves in sequence tell a complicated story:

1. Drop the safety pledge (February 2026): Remove the mechanism that could force an internal slowdown.
2. Withhold Mythos (April 2026): Make a genuine safety-first decision that costs revenue and draws praise from safety advocates.
3. Raise $65B and prepare for IPO (May–June 2026): Accelerate commercial scale at unprecedented speed.
4. Call for a global pause (June 2026): Ask everyone else to slow down.

Skeptics see a pattern of a company using safety advocacy to lock in a competitive advantage . Supporters see a company genuinely torn between its founding mission and the realities of competing in a capital-intensive industry where pausing unilaterally means losing. Both readings are plausible, and the available evidence does not resolve the ambiguity.

What is unambiguous is that the conflict is now structural. A company cannot simultaneously be the loudest voice calling for a global slowdown and the fastest-moving commercial player in the same race without facing a credibility problem. The market has noticed. The activists have noticed . And the IPO prospectus — when it becomes public — will force the company to reconcile the tension in language that carries legal liability.

Anthropic’s 2026 has made one thing clearer than ever: safety advocacy and commercial ambition can coexist in a mission statement, but they are increasingly difficult to reconcile on a balance sheet.