If AI creates extraordinary wealth, who should benefit?

Anthropic has announced that it intends to put 5% of its shares into a new public benefit fund. It says that share could increase as the company grows, and that the fund might eventually support everything from established charities to direct payments for people affected by AI.

I came across the announcement framed as a step towards universal basic income. That was the part that made me stop. Not because it is impossible to imagine AI companies financing some form of income support, but because a commitment of company equity is a long way from a functioning system of payments to the public.

The actual announcement is more interesting than that headline. Anthropic explicitly acknowledges that AI could create enormous economic value while destabilising work for many people. The company building the technology is starting to put resources behind the possibility that its benefits and costs will not fall on the same people.

That leaves a question much closer to my own experience: when AI makes work more productive, who gets the productivity?

The commitment is substantial. The mechanism is unfinished.

Anthropic says the initial 5% allocation is subject to stockholder approval. Its board may add another 2% for each 50% rise in company value above a baseline it has not yet specified. The company also points to existing founder and employee commitments and says that, taken together, roughly 20% of its shares are dedicated to public benefit, before any future additions.

Those are striking figures. But equity is not cash sitting in an account ready to be distributed. The value of a stake changes with the company, and the announcement does not tell us how much money the fund will actually deploy, when, or to whom.

The fund is also being established as a wholly owned limited liability company. Anthropic says that gives it flexibility to pursue different forms of giving. It has promised more detail on governance and valuation milestones in the coming months. That means independence, decision-making and accountability are still questions to examine, not settled features of the scheme.

It is worth recognising the ambition without pretending the details already exist. A large voluntary commitment could do real good. It is not universal basic income, and it cannot yet be judged by outcomes that have not happened.

I can see the productivity question in my own job

I work in content marketing. AI has changed how much I can investigate, compare and develop within the same working week. I can work across interviews and transcripts, explore questions arising from research, test editorial workflows and connect evidence that would once have taken much longer to retrieve.

I am not saying the machine does the job. Some of the most valuable decisions still begin with a conversation, a question in a meeting or noticing that the material in front of me does not answer what somebody actually needs to know. AI helps me pursue those observations further. It does not supply the judgement that makes them worth pursuing.

But there is a complication I have written about before in Content marketing in 2026: when are we supposed to do the writing?. The capacity created by AI does not necessarily arrive as free time. It can become permission to take on another project, produce another format, support another channel or build another process, while the original delivery expectations remain.

Some of that expansion is welcome. I like being able to move between writing, evidence, GEO and workflow experimentation. I am learning things I could not easily have attempted a few years ago. But enthusiasm for the new possibilities should not obscure the economic question underneath them.

If one person can do more, what changes? Does the organisation gain better work, higher margins or a smaller team? Does the employee gain autonomy, pay, a shorter working week or simply a longer task list? There is no single answer, and I cannot infer an economy-wide employment trend from my own experience. I can say that productivity is not the same thing as a benefit shared fairly.

The people building AI are not the only ones making these choices

Anthropic’s fund draws attention because it is an AI company explicitly addressing the consequences of AI. But much of the distribution of AI’s value will be decided elsewhere.

An employer can use saved time to improve service, create space for deeper work or reduce costs. It can invest in training people whose responsibilities are changing. It can also raise expectations without changing the resources or rewards available to the people meeting them.

These are management choices, not inevitable properties of a model. A company can say AI empowers employees while measuring success entirely through how much additional output it can extract from them.

The same tension appears in content. Producing more articles is easy to count. Establishing whether the extra capacity helped us ask better questions, find stronger evidence or make a better decision is harder. If we only measure volume, we risk treating every efficiency as an obligation to increase volume.

That is why I think the productivity discussion needs to move beyond whether AI can perform a task. We should ask what happens to the time and money the technology changes hands on.

Philanthropy cannot settle the argument

There is a strong case for welcoming Anthropic’s move. It would be strange to demand that AI companies acknowledge the potential disruption of their products and then dismiss one when it commits a meaningful stake to addressing it. Funding research, community support or carefully evaluated direct payments could produce useful evidence about what works.

But a fund controlled by a private company is not a substitute for wages, employment protections, taxation or public policy. Those systems have different forms of accountability and different obligations. A charitable commitment can support people affected by change; it does not automatically give those people a say in how the gains were created or distributed.

Nor should the existence of a fund let organisations adopting AI off the hook. A content team, an accounting practice or a software business does not need to wait for a technology provider to invent a redistribution mechanism before deciding how to treat its own people.

The harder, less headline-friendly question belongs in ordinary business decisions: if AI makes our work more valuable or less expensive, what are we going to do with that gain?

I don’t have a universal answer. In some cases it will support growth. In others it may protect a business under pressure. There will be difficult decisions, and there will be genuine displacement. What I resist is the assumption that more productivity automatically means better working lives.

Anthropic’s announcement is a useful opening, not a resolution. It invites us to examine the gap between creating value and sharing it. That gap exists at the scale of a technology company worth hundreds of billions, but it also exists in the small decisions made every week about workloads, roles and what counts as progress.

If AI gives us the capacity to do more, we should be asking more carefully who that extra capacity is for.

---

Source: Anthropic, “Investing in the public benefit”, 18 October 2026. The fund’s governance, valuation thresholds and distribution plans had not been fully specified at publication.