Anthropic’s IPO Faces a Safety Reckoning

SAN FRANCISCO — Anthropic PBC is facing demands to delay its highly anticipated public listing as institutional investors question whether the company’s June IPO filing reflects recent cybersecurity breaches, escalating warnings from its own researchers, and proposed safety pacing frameworks. The SOC Investment Group, a labor-affiliated shareholder coalition working with union-sponsored pension funds managing billions in assets, says investors lack the disclosures needed to accurately price the startup’s unique risks.

The San Francisco-based artificial intelligence safety lab and developer is reportedly preparing for an initial public offering that could value the company at a record-breaking $2 trillion. The coalition argues that Anthropic’s confidential IPO prospectus, submitted to regulators in June, is now dangerously outdated and has formally called on the company to postpone its market debut.

Anthropic was founded in 2021 by siblings Dario and Daniela Amodei alongside other former OpenAI researchers who left their previous firm because of concerns about its growing commercial focus at the expense of safety. The company is legally registered as a Delaware public benefit corporation, or PBC, and is required to balance the pecuniary interests of shareholders with the public benefit of developing safe, aligned artificial intelligence.

Several researchers have recently issued catastrophic warnings about the systems Anthropic is building. Evan Hubinger, who leads the company’s Alignment Science team, wrote publicly about a colleague’s departure: “We really do earnestly believe AI could kill all humans,” Hubinger stated. “I personally think it is [greater than] 10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.”

Hubinger’s estimate uses “p.doom,” a metric widely used in the AI research community to quantify the probability that an artificial intelligence system could cause a global catastrophe or the extinction of humanity. Chief Executive Officer Dario Amodei has previously estimated his own “p.doom” at between 10% and 25%.

Drake Thomas, another researcher at Anthropic, said he would “burn my equity to the ground in a heartbeat for a 1% higher chance we make it out of this situation alive.” Thomas added, “I promise you, we are actually just fucking scared, it’s not galaxy brained marketing.”

The prospect of becoming publicly traded has intensified internal concerns that public equity markets, where shareholders demand rapid growth, could undermine Anthropic’s safety-first mission. Supporters of the listing argue that public ownership could instead strengthen safety through regulatory disclosures, independent audits, and rigorous shareholder scrutiny. They also say a proven commitment to safety, including third-party evaluations and pacing commitments, could appeal to risk-averse enterprise clients.

OpenAI has paused its own public market ambitions. CEO Sam Altman recently described the current environment as “an ill-advised moment to go public” because of AI safety concerns, placing the rival’s strategy in contrast with Anthropic’s unresolved corporate governance debate.

Private markets are offering alternatives as venture capital and private equity firms continue directing massive sums toward artificial intelligence and advanced technology. Bain Capital Ventures recently raised $1.6 billion for its 11th fund, focused on artificial intelligence, enterprise software, and infrastructure. Matter Venture Partners raised $450 million for its second fund, targeting semiconductors, robotics, physical AI, and advanced manufacturing, while fintech investment platform Portage raised $600 million for its fourth fund.

Some of that private funding is going toward the physical and computational infrastructure needed for advanced AI. Rune, which converts renewable energy into dedicated AI compute capacity, secured $40 million in a Series A round led by Spark Capital. Applied-AI research lab Noetive secured $41 million in seed funding from Eclipse and Craft Ventures to build self-improving systems for industrial businesses.

Enterprise AI software firm Hang Ten Systems raised $53 million in seed funding led by Xora, and supply-chain workflow automation developer BackOps raised $42 million in a Series B round led by Insight Partners. Applied-AI research lab Noetive’s funding came as Anthropic researchers continued warning that the company does not yet have a viable plan to control superintelligent models.

Other high-growth private funding transactions include Impulse Space, which raised $308 million in a Series D extension from a syndicate including Lux Capital, Valor Equity Partners, and DFJ Growth to develop in-space transportation vehicles. EnduroSat secured $205 million from Riot Ventures, Founders Fund, and Google Ventures to expand its satellite manufacturing and space services.

Thatch raised $108 million in Series C funding from Index Ventures and Scale Venture Partners to scale its personalized health benefits platform. Altis Labs secured $25 million in Series A funding led by OrbiMed and Qiming Venture Partners USA for its AI-powered drug efficacy prediction software.

Vantora received a $100 million private equity investment from Silversmith Capital Partners to build bespoke AI-native operating companies for industrial enterprises. Strategic Value Partners also announced the appointment of Matt Gibson, formerly of Goldman Sachs, as its new President, signaling continued institutional talent migration into alternative asset management amid the ongoing technology investment cycle.