SAMPLEDEffective altruism, explained for business — and what it says about AI in music
What effective altruism actually is as a business framework, which people in and around music already operate this way, why the tech world adopted it so hard, and how the same logic lands on AI, record labels and the value of a song.
Effective altruism gets talked about as a philosophy, a subculture, or a punchline about a crypto exchange. In business terms it is much duller than any of that. It is a resource-allocation framework: given finite money, time and attention, which use of them produces the largest verified improvement per unit spent?
That is a budgeting question. Every founder already answers a version of it every quarter. Effective altruism just insists you answer it out loud, in writing, with numbers you are willing to be wrong about in public.
This is the third piece in our series. The first was effective altruism for creative entrepreneurs; the second was music charity, audited. This one goes wider: the business logic, the people, the tech-industry adoption, and the collision with AI that is currently rewriting how music gets valued.
The framework, in plain business language
The movement's operating screen has four parts, usually shortened to scale, neglectedness, tractability and evidence. Translated into language a label, a studio or a two-person startup would recognise:
- Scale — how big is the addressable problem? Total market, total harm, total people affected.
- Neglectedness — how crowded is the field? Crowded markets have thin margins on impact for the same reason they have thin margins on price.
- Tractability — can money and effort actually move it? Rent is tractable. Structural royalty reform is not, at least not on a one-year budget.
- Evidence — after the fact, what number tells you it worked? If nobody can name one before you spend, nobody will name one after.
The most business-shaped part of it is the counterfactual. Not "did good things happen?" but "what would have happened anyway?" That is the same question a marketing director should be asking about a campaign and usually is not.
The research infrastructure around this is public and unusually rigorous. GiveWell (opens in a new tab) publishes its cost-effectiveness spreadsheets, including the mistakes. Open Philanthropy (opens in a new tab), funded largely by Facebook co-founder Dustin Moskovitz and Cari Tuna, publishes its grant reasoning. 80,000 Hours (opens in a new tab) applies the same analysis to careers rather than cash — which is the version most relevant to anyone building a company instead of writing cheques. Founders Pledge (opens in a new tab) exists specifically for founders who want to commit a share of an exit before there is an exit to argue about, and Giving What We Can (opens in a new tab) runs the 10% income pledge that gave the movement its early shape.
For a business, the practical yield is three habits: pick one metric that would embarrass you if it stalled, run the counterfactual before you commit spend, and publish the result even when it is unflattering. None of that requires believing anything philosophical.
The critiques are part of the framework
Any honest version of this includes the objections, because the movement's own tools generate them.
Measurement bias. What gets counted gets funded. Interventions with clean randomised evidence outrank interventions that matter but resist measurement — organising, advocacy, culture. Music is almost entirely the second category.
Longtermism drift. A large share of EA money moved from bed nets to speculative long-horizon risk, especially AI. Reasonable people think that is either the most important reallocation of the century or a very well-argued way to fund your friends' research.
FTX. Sam Bankman-Fried built a public identity around "earning to give" and was convicted of fraud in 2023. The movement's response — and the reason the critique bites — is that a framework built on verification failed to verify its most visible donor. Any founder borrowing the vocabulary should borrow that lesson first: governance is not a rounding error on impact.
Who in music actually operates this way
Very few people in music say the words "effective altruism." A number of them run something structurally close: a fixed share of revenue, a named outcome, an outside body checking the number.
Brian Eno is the clearest case. He co-founded EarthPercent (opens in a new tab), which asks artists, labels and companies to commit a small percentage of income to climate organisations chosen by an advisory panel rather than by whoever posted the most compelling appeal. Percentage-of-revenue giving plus delegated grantmaking to people who study the field is, functionally, the pledge model.
Massive Attack did something rarer than donating: they commissioned research. The band worked with the University of Manchester's Tyndall Centre for Climate Change Research (opens in a new tab) on a study of live music's emissions, then staged a show built to test the findings. That is evidence-first philanthropy — pay for the measurement before you claim the outcome.
Coldplay published tour sustainability figures with an academic reviewer attached, an approach documented on the band's sustainability page (opens in a new tab). Agree or disagree with the totals; the disclosure format is the point, and it is more than most touring businesses of that size offer.
Moby has given away restaurant profits and campaigned on animal welfare for decades, drawing openly on Peter Singer's arguments — Singer being the philosopher whose work seeded effective altruism in the first place. The lineage is direct even where the label is never used.
Bono built ONE (opens in a new tab) and (RED) around data-led advocacy rather than direct relief, an approach argued over constantly and never quietly. Whatever you make of it, it is a deliberate bet on tractability over visibility.
The pattern across all of them: a defined share of income, a third party choosing or checking the destination, and published numbers. That is the transferable part. It works for a mid-size label as well as it works for a stadium act.
Why the tech industry adopted it so hard
Effective altruism found its money in software for reasons that are not mysterious. The framework rewards people with volatile, concentrated wealth, quantitative training and no inherited philanthropic institutions to inherit. Software produced all three at once.
Open Philanthropy's scale came from Moskovitz and Tuna. Much of the AI safety research agenda — and, by extension, a large slice of the people now building frontier models — grew out of the same donor network and the same argument about neglected long-horizon risk. Anthropic, OpenAI and DeepMind all recruited from a talent pool that had been reading the same essays for a decade.
Which is the part that matters for musicians: the people arguing about how AI should be governed and the people building the AI that is currently ingesting recorded music are, to a meaningful extent, the same community. Understanding how they reason is not academic. It is opposition research.
Now connect it to AI and music
Apply the four questions to AI's arrival in music and you get an unusually clear read on where the industry's fight actually is.
Scale. Generative audio touches every revenue line at once — recording, publishing, sync, session work, production services. The international authors' body CISAC commissioned a study with PMP Strategy projecting that a substantial share of music creators' revenues is at risk of erosion by 2028 as generative outputs compete with human work; the CISAC study (opens in a new tab) remains the most-cited attempt to put a number on it. Cross-reference it with the Stanford HAI AI Index (opens in a new tab) for the capability curve driving those projections.
Neglectedness. Almost nobody is funding the boring middle. Litigation is funded, lobbying is funded, model development is very funded. Metadata infrastructure, consent registries, provenance standards and the ability of an independent artist to prove what their voice is worth — largely unfunded. The Coalition for Content Provenance and Authenticity (opens in a new tab) is doing the standards work; almost none of the money in this argument is going there.
Tractability. Licensing moves faster than legislation, which is why the majors have gone to the negotiating table rather than only to court. Watch the primary sources rather than the coverage: Universal Music Group's newsroom (opens in a new tab), Sony Music's press page (opens in a new tab) and Warner Music Group's press releases (opens in a new tab) are where the actual terms surface first, and the deal structures they publish will set the default rate for everyone downstream. The industry body IFPI (opens in a new tab) tracks the aggregate picture in its Global Music Report; the RIAA (opens in a new tab) does the same for the US market.
Evidence. The honest answer is that nobody has clean numbers yet on how much AI-generated audio is displacing human streams versus expanding total listening. Anyone selling you certainty on that in 2026 is selling something. The measurable proxies available now are catalogue acquisition prices, session musician bookings, sync placement rates and the share of new uploads flagged as synthetic — all of which any working professional can watch without a research budget.
What this means if you make music for a living
The EA reading of AI in music is not "the robots are coming." It is narrower and more useful:
- The neglected asset is proof. Registered works, clean splits, consistent ISRCs and documented provenance are the things that will determine whether you get paid in a licensed-AI economy. Our music metadata guide covers the mechanics.
- Consent is a business decision, not a moral pose. Whether you allow your voice, your catalogue or your stems into training data is a licensing question with a price attached. Decide the price before someone else decides it for you.
- Counterfactual thinking applies to your own output too. If a model can generate the thing you spent the week on, the week was allocated badly. That is uncomfortable and it is also just capital allocation.
- Collective bargaining is the highest-leverage giving in music right now. Not because it is emotional, but because it is neglected and tractable at once — the two conditions the framework says to look for.
The Sampled position
We run this publication with a stated bias toward primary sources and coverage nobody else is doing, which is a neglectedness bet in the same vocabulary. We are not affiliated with any organisation named here, and we take no fees from them.
Effective altruism's actual contribution to business is not a cause list. It is the demand that you write down what you expect to happen, spend, and then go back and check. The music industry is about to make a series of very large, very permanent decisions about AI with almost none of that discipline attached. It would be worth borrowing.
Sources are linked inline. Sampled is independently operated.