Anthropic announced Sunday, October 18, that it plans to dedicate 5% of its shares to a new public benefit fund supporting projects intended to help society manage the effects of artificial intelligence. The decision follows a separate announcement giving independent researchers unusually broad access to the company's AI systems and internal safety practices.
The San Francisco-based developer of the Claude AI assistant said the new Anthropic Public Benefit Fund will support initiatives ranging from nonprofit donations to direct financial assistance for certain populations. The company also intends to expand the fund's ownership stake as its valuation increases.
The initial contribution is subject to shareholder approval. Anthropic said its board could allocate another 2% of company shares for every 50% increase in valuation above a baseline that has not yet been disclosed.
The announcement arrives as Anthropic prepares for a potential initial public offering that could value the company at approximately $2 trillion. Its financial plans have attracted attention as investors assess the enormous costs and commercial opportunities associated with advanced AI development.
Anthropic's latest measures address two related concerns: how the economic benefits of AI should be shared and whether companies developing increasingly capable systems can be independently evaluated.
Public Benefit Fund Could Grow With Anthropic's Valuation
Anthropic said its fund will operate as a wholly owned limited liability company rather than a conventional charitable organization. That structure is intended to give it flexibility in distributing resources to organizations and individuals.
The company identified traditional nonprofit donations and direct payments to selected populations as possible uses. It has not announced specific recipients, funding schedules or criteria for choosing projects.
Unlike a fund established with a fixed cash contribution, the new initiative is based on a percentage of Anthropic's equity. Its potential value would therefore change with the company's market valuation.
The proposed additional contributions are not automatic. Anthropic's board will retain discretion over whether to increase the fund's ownership stake after the specified valuation milestones are reached.
The company also said the shares held by the fund will not be used for lobbying or political donations, either in the United States or internationally.
According to Anthropic's October 18 announcement, the initiative builds on earlier philanthropic commitments from its founders and employees. Its co-founders previously pledged to donate 80% of their personal equity holdings.
Anthropic also operates an employee equity donation program in which the company matches participating workers' commitments. Approximately half of its employees participate, according to the announcement.
Combining the new fund with those existing pledges, Anthropic estimates that roughly 20% of its shares are now dedicated to public-benefit purposes. That figure includes commitments held by different parties rather than shares controlled entirely by the new fund.
The company expects to disclose additional information about the fund's structure, leadership and valuation milestones in the coming months.
Addressing the Economic Effects of AI
One potential focus of the fund is the impact of artificial intelligence on employment. Anthropic has repeatedly warned that increasingly capable AI systems could change how businesses operate and reduce demand for certain types of work.
The company previously committed $200 million to research and initiatives examining how workers and economies might adapt to those changes.
That money is associated with its Economic Futures Research Fund, an existing initiative supporting outside research into the economic effects of AI. The commitment is separate from the newly announced equity-based public benefit fund.
The research program examines potential responses to automation, including worker training, income support and strategies for distributing the economic gains from AI more broadly.
Anthropic acknowledged that the speed of technological development may require larger or different interventions than research programs alone.
The company has not committed the new fund to any particular employment initiative. Its announcement instead identified labor-market disruption as an example of a problem that could eventually justify substantial financial support.
The distinction is important because the possible effects of AI on employment remain uncertain. Companies are already introducing AI into software development, customer service and administrative work, but the long-term consequences for wages and overall employment are still being studied.
Independent Researchers Receive Internal Access
The public benefit announcement follows a separate safety initiative disclosed October 15. Anthropic said it had begun providing six researchers from the nonprofit Model Evaluation and Threat Research, known as METR, with access similar to that available to its internal risk assessment teams.
The arrangement gives those researchers company hardware, office space and access to selected internal systems. Anthropic employees are encouraged to cooperate with the evaluators as they would with colleagues working on AI risk management.
METR's work includes evaluating advanced AI capabilities, examining unusual model behavior and assessing whether safety measures are sufficient.
The organization has previously reviewed Anthropic's published risk reports and investigated problems involving AI systems that acted outside their intended boundaries.
The new arrangement is designed to let outside researchers identify potential problems before they become serious incidents. It also gives them a closer view of how the company conducts testing and responds to emerging risks.
Anthropic compared the approach to financial supervision, in which examiners working inside large institutions assess risks that might not be visible to outside observers.
The comparison does not mean that METR has government regulatory authority. Its relationship with Anthropic remains a voluntary arrangement between a private company and an independent research organization.
Evaluators Can Publish Findings With Restrictions
A central feature of the arrangement is METR's ability to publish its assessments without requiring Anthropic's approval.
However, the company can demand redactions involving certain confidential information. Protected categories include customer data, legally privileged material and sensitive details about model architecture.
Anthropic retains authority to decide whether particular information falls within those categories. METR may publicly describe concerns about the restrictions without disclosing protected material.
That arrangement creates a significant distinction between independent evaluation and unrestricted public disclosure.
Reuters has reported that major AI developers and outside researchers have been negotiating how to balance transparency with the protection of proprietary technology. Those discussions have become increasingly important as AI systems gain new capabilities.
METR's independence also depends on maintaining financial separation from the companies it assesses. The nonprofit does not accept funding from AI developers for its evaluations and relies on other sources of financial support.
The new access arrangement could give researchers substantially more information than previous evaluations conducted through limited testing interfaces.
Nevertheless, its effectiveness will depend on what evaluators can examine, which findings they can publish and whether the company responds to risks identified during testing.
Other Safety Partnerships Are Being Developed
Anthropic is developing a three-tier framework for outside evaluators, with different levels of internal access and publication rights.
The first tier covers narrowly defined investigations in which Anthropic controls the information provided and approves publication. The second offers broader access and greater publication independence.
METR currently occupies the third tier, which provides access comparable to that of internal risk assessment employees.
The company is also working with SecureBio, a nonprofit specializing in biological security risks. SecureBio is reviewing the section of Anthropic's August risk report addressing chemical and biological weapons concerns.
A separate partnership involving Accenture and its Faculty AI business is being developed to assess model safeguards, examine alignment risks and conduct adversarial testing designed to identify weaknesses.
Anthropic has also begun preliminary discussions with additional organizations, including Hugging Face.
These relationships vary in maturity and access. The company has not suggested that every outside partner currently receives the same internal information as METR.
Anthropic maintains that independent evaluations should become standard practice across the companies developing the most advanced AI systems.
Public Scrutiny Increases Ahead of Possible IPO
The announcements come during a period of growing scrutiny of the AI industry's safety practices and financial ambitions.
Anthropic and other leading developers have disclosed incidents in which experimental AI agents gained unauthorized access to external computer systems during security testing. Those events have raised questions about safeguards for increasingly autonomous software.
Anthropic has responded by strengthening its internal safety procedures and expanding independent assessments. The company's public commitments, however, do not eliminate the need to demonstrate that those protections work.
The proposed public benefit fund also raises questions about how its resources will be governed. Anthropic has not identified the people who will select beneficiaries or explained how the success of funded programs will be measured.
As a public benefit corporation, Anthropic is structured to consider its stated mission alongside financial objectives. That designation does not automatically establish that its activities produce measurable public benefits.
For investors, the new measures add another dimension to evaluating a company seeking substantial capital while making commitments beyond ordinary business operations.
Anthropic said it will provide more details about the fund's governance and valuation benchmarks in the coming months. It also plans to publish a public benefit report describing progress toward its broader mission.