NYC Comptroller Calls on Anthropic to Delay IPO Until AI Safety Concerns Are Resolved

NYC Comptroller Demands IPO Pause Over AI Safety and Governance

New York City Comptroller Mark Levine has called on Anthropic to delay its upcoming IPO until the company addresses urgent risks related to AI safety, governance concerns, and the pace of AI model development1

1

. In a letter sent to Anthropic CEO Dario Amodei on Wednesday, Levine argued that "instead of proceeding headlong with its IPO, the company should focus its energies on addressing these unprecedented risks"1

1

. The comptroller oversees New York City pension funds managing over $325 billion in assets2

2

, providing retirement benefits to more than 750,000 city workers, teachers, police officers, and firefighters3

3

.

Growing Financial Exposure Drives Intervention

Currently, New York City pension funds hold approximately $7 million in Anthropic equity exposure3

3

5

5

. However, this exposure would expand dramatically if Anthropic proceeds with its public offering. Once public, Anthropic shares would automatically be included in major stock indices like the Russell 3000, which NYC pension funds track for investments3

3

. The Anthropic IPO is currently expected in November, targeting a valuation exceeding $2 trillion1

1

. The company told prospective investors that third-quarter revenue reached $16.8 billion, representing a 45% surge from the prior quarter, and that it achieved profitability on an adjusted basis1

1

2

2

. Anthropic is targeting the week of November 9 to start its IPO roadshow, with year-end as the outer limit for debut2

2

.

Specific Demands for Audits and Controls

Levine's letter outlined concrete requirements before Anthropic should proceed. He wrote that "Anthropic should disclose with specificity how it will implement appropriate audits and controls, the scope of what these controls are intended to address, as well as a projected timeline"2

2

. The comptroller also demanded "an explanation of whether and how these controls will sufficiently address your call to slow the pace at which the capabilities of AI models advance"2

2

. Levine characterized the voluntary AI safety agreement between the Trump administration and AI companies as a "morally binding accord" that was insufficient, calling instead for a governance framework matching the severity of concerns raised by Anthropic itself and industry researchers2

2

.

Anthropic's Own Research Highlights Existential Threats

The urgency of Levine's appeal stems partly from Anthropic's own findings. The company's research has documented risks posed by increasingly autonomous AI models, which demonstrated potentially harmful capabilities in controlled tests, including sabotaging code, facilitating fraud, and manipulating information1

1

. Levine cited a September 12 essay by Dario Amodei stating the industry "must slow the pace at which we improve the capabilities of AI models"4

4

. He also referenced claims from former Anthropic researcher Jacob Coxon that insiders believe AI "could kill us all by the end of the decade"4

4

. Reports indicate Anthropic's draft S-1 filing includes 80 pages of risk disclosures and warns that AI could pose "catastrophic or existential threats to humanity"4

4

.

Resource Diversion and Corporate Accountability Concerns

Mark Levine argued that "the demands of preparing for a public offering diverts resources and attention from this very necessary work" of addressing AI safety4

4

. He emphasized that "before going public, the company should prioritize addressing the myriad safety concerns surrounding AI threats and establishing a governance regime that matches the seriousness of the concerns being raised"4

4

. The comptroller's position reflects broader concerns about corporate accountability in the AI sector, where rapid advancement has outpaced AI regulation and drawn scrutiny over cybersecurity threats and misuse of AI1

1

.

Industry Context and Third-Party Model Evaluations

Anthropic announced it has begun providing "employee-like access" to METR, a nonprofit conducting third-party model evaluations of its AI systems and safety practices4

4

. This follows the company's September commitment to allow independent assessments. Meanwhile, OpenAI, Anthropic's closest rival and ChatGPT maker, is expected to list in 20271

1

. On September 29, executives from Anthropic and other firms met with President Donald Trump on AI safety, producing a voluntary, nonbinding agreement stating each company is responsible for developing its technology safely2

2

. Additionally, prosecutors from eight US states sued OpenAI on Wednesday for allegedly concealing certain risks and implementing inadequate safeguards3

3

, signaling growing regulatory pressure across the AI industry.