The honest view from inside technical leadership
Analysis

Engineering leadership

Independent reporting on the decisions, costs and contracts that shape how engineering organizations build with AI.

AI Governance · News

Anthropic’s AI Pacing Pledge at Its IPO Depends on Evaluators It Cannot Veto

Dario Amodei told the Nasdaq opening ceremony that Anthropic will slow the training and release of its models whenever safety requires it. The people positioned to tell the public whether that happens are six METR staff working inside the company.

Line-art illustration of a stock ticker line passing through a gate held by a magnifying glass, representing Anthropic’s AI pacing pledge and the independent evaluators who check it

Anthropic CEO Dario Amodei pledged on 19 November 2026, in a speech ahead of the company’s first trade on Nasdaq, that Anthropic will pace the training and deployment of its AI systems for as long as it takes to be confident they are safe and under human control. The outside check on that AI pacing pledge is a team from the nonprofit evaluator METR, which works inside Anthropic and can publish findings Anthropic says it cannot veto.

Amodei Made the AI Pacing Pledge in Anthropic’s Nasdaq Opening Bell Speech

Amodei read the speech from a printout before the opening bell, TechCrunch’s Sarah Perez reported, and the outlet published the full text. He opened with the loss of his father to a disease that became curable a few years later, and with the case that AI can speed up medical research. He then listed the risks he said Anthropic has always acknowledged, from misuse by individuals or nation-states to economic disruption to losing control of AI systems, and called them avoidable if the builders of AI choose to make them so. The commitment came near the end. Anthropic, he said, will pace training and deployment for as long as necessary, and “we will not bend to either business nor political pressure.” He argued that the money raised that day strengthened the company’s resolve, and that showing safety and commercial success together would push competitors to follow, which he called a race to the top.

Anthropic Raised About $81.3 Billion the Day Before the Pledge

Anthropic, PBC priced its offering on 18 November 2026 at $83 per Class A share for 980,000,000 shares, about $81.3 billion, according to the company’s pricing announcement. Of those shares, 956,351,354 were sold by Anthropic and the rest by existing stockholders, and underwriters received a 30-day option on 147,000,000 more. The stock began trading on the Nasdaq Global Select Market under the ticker ANT. CNBC reported that it opened at $111.50, closed its first day at $97.17, and left Anthropic with a fully diluted market value of roughly $1.8 trillion, in what CNBC described as the largest IPO on record. Every buyer of those shares now owns a company whose CEO has told them, on the first day, that releases will wait on safety work.

Amodei First Proposed Pacing AI Development in September 2026

The IPO speech repeated a position Amodei set out in September 2026, when he proposed pacing AI development through coordination among leading companies on safety standards and through third-party evaluators embedded inside AI labs, TechCrunch reported. OpenAI’s Sam Altman, Google DeepMind’s Demis Hassabis and xAI’s Elon Musk said they agreed with him. TechCrunch notes that few concrete commitments followed. A pledge to slow down has no measurable content on its own, because only the company sees the internal evaluations that would justify a delay. Whether the promise means anything depends on who else gets to look.

METR Can Publish Its Findings on Anthropic Without Anthropic’s Approval

Anthropic described the answer in a post dated 15 October 2026. It has onboarded six METR employees with company hardware, office space and access to internal tools and systems that it says are roughly on par with its own risk assessment teams. METR will publish its analysis of Anthropic’s risk levels, incidents and practices, and Anthropic says it cannot veto publication. The limit is redaction. Anthropic can require METR to remove information in predefined categories, such as confidential intellectual property, and Anthropic decides what falls inside those categories, while METR keeps the right to state how much access and information it received. METR sits in the top of three access tiers Anthropic has defined. First-tier evaluators publish only with Anthropic’s approval, and second-tier evaluators publish without it. SecureBio is reviewing the chemical and biological weapons chapter of Anthropic’s August Risk Report, and the company says it is working toward a broader evaluation partnership with Accenture and Faculty. If the pacing pledge were broken, a METR publication, or a METR statement that its access had narrowed, is where outsiders would be positioned to see it.

Why This Matters for Engineering Leadership

Anthropic has put a release gate in front of public shareholders and attached an outside reader with the right to publish. That structure is the useful part, and it transfers to any engineering organization that makes promises about when it will ship. A commitment such as “we do not release until the security review passes” is checkable only when someone outside the delivery team can see the evidence and report on it without the delivery team’s sign-off, and when the limits on what they may say are written down in advance. Engineering leaders can apply the same test to their own gates: name the reviewer, define what they can see, and decide ahead of time what they may publish internally and who can redact it. The pledge also changes vendor planning for teams that build on Claude. Anthropic’s CEO has now told investors that deployment dates bend to safety evaluations, so roadmaps should be built on models that are already released, with METR’s published reports read as part of vendor due diligence.

FAQ

What did Dario Amodei pledge at Anthropic’s IPO?

In his Nasdaq opening bell speech on 19 November 2026, Amodei said Anthropic will pace the training and deployment of its AI systems for as long as needed to be confident they are safe and under human control. He said the company will not bend to business or political pressure.

How much did Anthropic raise in its IPO?

Anthropic priced 980 million Class A shares at $83 each, about $81.3 billion in total, and began trading on Nasdaq under the ticker ANT on 19 November 2026. CNBC reported the shares opened at $111.50 and closed at $97.17.

Who checks whether Anthropic keeps its AI pacing pledge?

Anthropic has embedded six staff from the nonprofit evaluator METR with access to internal tools roughly on par with its own risk teams. METR will publish its analysis, and Anthropic says it cannot veto publication.

Can Anthropic edit what METR publishes?

Anthropic can require redactions in predefined categories, such as confidential intellectual property, and it decides what falls into those categories. METR keeps the right to say how much access and information it received.

When did Anthropic first propose pacing AI development?

Amodei set out a plan to pace AI development in September 2026, including industry coordination on safety standards and embedded third-party evaluators. TechCrunch reports that other AI leaders voiced agreement, though few concrete commitments followed.

Source: At Nasdaq opening bell, Anthropic CEO recommits to pacing AI development by Sarah Perez, TechCrunch, November 19, 2026. Primary sources: Anthropic, “Anthropic prices initial public offering”, November 18, 2026, and Anthropic, “An update on independent evaluators”, October 15, 2026. Trading figures: CNBC, November 19, 2026.

Read more news & analysis