Anthropic sets up a fund to compensate those who lose their jobs due to AI

Anthropic’s announcement of the Anthropic Public Benefit Fund, to which it will allocate 5 per cent of its shares, might seem like yet another philanthropic – and pre-IPO marketing – exercise from Silicon Valley. Yet the introductory post contains a detail that warrants closer examination: amongst the possible uses, alongside donations to non-profit organisations, Anthropic specifically mentions ‘direct distributions to specific populations’. In other words: money paid directly to people. And the first area mentioned is employment – specifically, those at risk of losing their jobs as a result of artificial intelligence.

The figures help to paint a clearer picture: Anthropic is preparing for a listing on Wall Street that could take place as early as November, and analysts had estimated a valuation of over two thousand billion dollars, with a five per cent stake worth around one hundred billion. Last night, however, the Wall Street Journal revealed a loss of 80.7 billion in the first half of the year – largely an accounting loss – and some investors are now estimating a valuation of around 1,500 billion. The board of directors will then be able to add two per cent for every fifty per cent increase in value above a threshold yet to be set: the more the company grows, the more the share allocated to the public good increases. The fund will be a wholly-owned limited liability company, a structure that frees it from the constraints of charitable foundations and allows it to make payments to individuals.

In June, Dario Amodei had written that AI could lead to lasting job losses, linked to the very nature of the technology. The company had accompanied the essay with a framework of graduated responses, which, in the worst-case scenario, envisaged a universal basic income funded by taxes on AI companies or on capital gains. The message was aimed at Washington, and it went something like this: tax us.

Washington has dropped the proposal. The Trump Administration has built its AI policy around the race with China, dismantling regulations and accelerating the roll-out of data centres; for workers, the 2025 action plan primarily provides for funding from the Department of Labour for reskilling. Meanwhile, relations between Anthropic and the White House have soured, to the point where its products have been formally removed from federal offices. In this climate, a tax on AI companies is the stuff of political fiction.

This is how Anthropic built the tool itself. The fund announced on Sunday is the first building block of a basic income paid for out of profits from artificial intelligence – the very same idea that Amodei was calling on the state to implement through the tax system. The closest comparison is the Alaska Permanent Fund, which since 1982 has paid each resident an annual dividend derived from oil revenues, with no other conditions attached: a public entitlement linked to residence, and therefore a basic income in the technical sense. If Anthropic’s fund were to pay only those who have been replaced by a machine, strictly speaking it would fall into a different category – that of unemployment benefit – much closer to the Five Star Movement’s universal basic income than to the universal models favoured by economists. And it would be the board of directors of a listed company that would determine who falls within the category of ‘specific populations’.

This is precisely where the risk lies. American history is well acquainted with philanthropy stepping in to fill the gaps left by the state: the Carnegie and Rockefeller foundations were established at a time when the federal government was failing to provide for entire areas of social welfare, and they ended up shaping healthcare and education for decades in line with their founders’ priorities. In Italy, the closest example is Adriano Olivetti, who, in Ivrea, provided his employees with services that the state did not yet offer. These are virtuous models whilst they last, but they are dependent on the will of those who fund them and the health of their finances. A basic income linked to a stock market share shrinks when the share price falls – that is, precisely when it is needed most.

Amodei has described the growing mistrust of AI as a “crisis of confidence”, and the fund comes just a few weeks before the listing, when confidence is reflected in the market, particularly in the share price. The fact remains that one of the companies transforming the world of work has set aside a portion of its capital for those who will pay the price. Washington might interpret this as an excuse to continue turning a blind eye. That would be the wrong interpretation: the fund functions as a prototype, and prototypes serve to demonstrate that something can be done, whilst we wait for someone with greater legitimacy to make it universal.