• Anthropic CEO Dario Amodei says AI poses "very real risks" and calls for pacing frontier-model development, even as the company prepares for a potential blockbuster IPO.
  • A September 29 White House meeting with President Trump could shape whether Washington favors voluntary safety reviews or a lighter, competitiveness-first approach.
  • Internal dissent and a looming public listing add pressure on Anthropic to balance mission-driven governance with investor expectations.

Safety Push Meets Market Reality

Anthropic CEO Dario Amodei has intensified his warnings that artificial intelligence carries "very real risks," arguing that frontier labs must slow the pace of capability breakthroughs so safety work can catch up. His proposal, outlined on September 12, centers on independent evaluators embedded inside leading AI labs, voluntary coordination among firms in democratic countries, and eventual international rules.

Amodei identifies the dangers as loss of human control, cyber misuse, bioterrorism, and broad economic disruption. At the United Nations in late September, he and OpenAI CEO Sam Altman urged international cooperation on AI safeguards. Amodei said Anthropic would slow releases as necessary to ensure successive systems are safe.

That message is colliding with the company’s commercial trajectory. Anthropic is considering a U.S. IPO and has projected roughly $190–$200 billion in 2028 revenue, according to Reuters. The company raised $30 billion in February 2026 at a $380 billion valuation, then $65 billion in May at a $965 billion post-money valuation. A Reuters-reviewed prospectus reportedly shows $4.6 billion of 2025 revenue and a $42 billion loss that year — a stark reminder of the cost of training and serving frontier models.

Pressure is coming from inside as well. Former Anthropic researcher Jacob Coxon resigned, arguing that frontier labs were developing systems they may not be able to control. His departure helped prompt Amodei’s public call for pacing.

The central policy fight is whether safety oversight can be strengthened without weakening U.S. competitiveness against China. President Donald Trump is scheduled to meet Amodei and other technology leaders on September 29, with reporting indicating the discussion is occurring amid mounting concern over AI-agent behavior and disagreements over stricter regulation.

A Governance Experiment Before the Bell

Anthropic has taken an unusual step before its planned IPO: its filing describes a "Founder LLC" through which seven founders would control a Class F share carrying 50.1% of voting power on key corporate matters. The company would remain a Delaware public-benefit corporation, formally allowing leaders to weigh public interests alongside shareholder returns. It is a structure designed to preserve mission influence after listing — and one that may raise governance questions for prospective public investors.

Founded in 2021 by former OpenAI leaders and researchers, Anthropic has become a major frontier player with its Claude family of models, APIs, and enterprise offerings. Daniela Amodei is president and board chair. The company’s Responsible Scaling Policy, initially released in 2023 and updated in August 2026, formalizes its approach to safety.

In June, Anthropic proposed stronger government authority over catastrophic-risk models. The White House’s AI Action Plan, by contrast, emphasizes removing rules that “unnecessarily hinder” development. An executive-order framework reportedly asks developers to submit certain covered frontier models for government review up to 30 days before release to trusted partners.

Critics argue that regulation can be hard to design, susceptible to regulatory capture, and may hand an advantage to less-regulated foreign competitors. Amodei has countered that voluntary promises are insufficient because market incentives reward faster capability gains.

The outcome of the September 29 meeting could clarify whether Washington pursues a voluntary review framework, formal legislation, or a lighter-touch approach. For Anthropic, the stakes are twofold: its IPO will put revenues, losses, infrastructure commitments, and governance under greater scrutiny, while its safety advocacy will be tested by the same competitive pressures it warns against.