An open letter to the open source community

We believe open source needs to evolve.

For decades, open source has given people a reason to believe that software could be built differently. You could learn from someone else’s code, repair what you depended on, and contribute an improvement that others could use in turn. People separated by geography or institutional boundaries could work together. A small team could build on tools it could never have afforded to create alone.

That promise shared the optimism of the early internet: wider access to knowledge would let more people participate, and cooperation could loosen the hold of established gatekeepers. John Perry Barlow’s 1996 declaration imagined an online world where economic power and inherited privilege would no longer determine who could enter. Opening access seemed a way to build a more equal world.

Open source gave that hope practical force. The ability to inspect, adapt, and share code helped people learn, build independent tools, and maintain alternatives to proprietary systems. Its usefulness did not depend on everyone having the same motives. People and businesses could cooperate because each could build on the result. These are lasting achievements, and we want to carry them forward.

Removing barriers to use made sense in that context. Each permission someone no longer had to seek made it easier to participate. Permissive licenses extended that approach with few obligations on reuse, while copyleft used obligations to preserve those freedoms when modified versions were redistributed. The open-source campaign launched in 1998 also made a practical case to business: participation in open development could produce useful software and commercial value. Wider adoption offered a way for shared work to reach more people.

Equal permission is not equal power

The Open Source Initiative’s definition protects the ability to use, modify, and share software across people, groups, and fields of endeavor. Those permissions let people cooperate across different interests and make it easier to build on one another’s work.

The difficulty begins when we treat those permissions as a sufficient answer to the larger promise. Removing a barrier to participation does not determine who has the resources to benefit most, or who controls what is built. When a license places no conditions on an exploitative use, its terms provide no way to challenge that use. Openness alone leaves those harms to be addressed elsewhere.

The definition governs software permissions. It does not require equitable compensation for employees, a share of profits for contributors, democratic control of a workplace, or funding for ongoing maintenance. It leaves those economic relationships outside its criteria. Nondiscrimination in licensing does not, by itself, produce equality in economic life. Equal permission to use software does not mean equal power to benefit from it.

An independent developer, a worker cooperative, and a multinational corporation can receive the same permissions while having profoundly different power to turn them into wealth. Capital, infrastructure, market access, and ownership determine who can operate at scale and who controls the proceeds.

We want the cooperation and independence that drew people to open source to extend to the people doing the work. That requires us to judge how its benefits are distributed, as well as how freely its code can be used.

We reject the premise that equal permission alone ensures equitable freedom. Software freedom should account for the people who create its value—and the power of those who capture it.

How shared work can concentrate wealth and power

When a business incorporates shared software, it can reduce the cost of building its products and running its operations. Contributors have supplied part of that value. Employees supply more through the work of adapting, maintaining, and using it. Neither group is guaranteed a share of the resulting profits or a say in their distribution.

Where those gains are retained by a small group of owners and executives, they can use them to expand infrastructure, acquire businesses, control distribution, and strengthen their position in the market. That stronger position can help them capture more of the value created by the next round of shared work. Freely contributed software becomes an input to a cycle that reinforces existing concentrations of wealth and power.

This is a consequence we should judge, even when everyone involved has complied with the software license. Permission to use someone’s work does not settle the question of what the beneficiary owes to others.

Open source also supports education, independence, public infrastructure, and alternatives to dominant firms. Those benefits are worth protecting. They do not make every commercial use equitable or ensure that its gains reach the people doing the work.

AI makes the stakes larger

AI helps more people create software, learn, and put ideas into practice. Wider access to those capabilities does not, however, democratize ownership of the models, computing infrastructure, or platforms through which that work is done.

The industry’s capabilities also rest on work contributed by others. Code-generation models such as Codex were trained on publicly available code. Datasets such as The Stack assemble source code at enormous scale for model training. That shared software embodies the labor of developers and communities, including work contributed without payment.

The unpaid labor embodied in those contributions can become training material for a commercial product without giving its contributors a corresponding share of the product’s ownership or income. We see extraction in an arrangement that takes the benefit of that labor while leaving responsibility for sustaining its creators optional. The fact that work was available under an open-source license does not mean its creators were compensated for the value it later helped produce.

AI can therefore broaden access to useful capabilities while reproducing the same unequal distribution of rewards. A 2025 IMF working paper models how AI can raise returns to capital and widen wealth inequality even while reducing some wage inequality. More productive tools do not automatically distribute the wealth they create more equitably.

We want AI’s benefits to reach more people. That means taking seriously the shared and unpaid labor behind it, the ownership structures around it, and who has the power to decide how its gains are used.

The cost is already being paid

These arrangements have costs in people’s time, livelihoods, and ability to keep contributing. In Tidelift’s 2024 maintainer survey, 47% of respondents reported receiving no pay for maintaining projects. In a separate question, 60% said they had quit or considered quitting maintenance work.

When contribution depends on having time to work without pay, the ability to participate is unequal too. A maintainer’s need to make a living does not disappear because others depend on their software. Projects can lose the people who understand them, and users can become dependent on code whose upkeep lacks adequate resources.

AI is also changing the pressure on software security. In its October 2026 research agenda, the UK’s National Cyber Security Centre describes evidence of AI lowering the barrier to some offensive tasks and increasing the speed and scale of others. Tools that make vulnerability research and exploitation easier increase the urgency of finding and repairing flaws.

The workload includes legitimate discoveries as well as attacks. In an October 2026 account, curl maintainer Daniel Stenberg described the project’s current vulnerability-reporting challenge as high volume and high quality. Better discovery still requires investigation, safe fixes, testing, coordination, and releases. AI can help with that work, but producing more findings does not itself provide the time or funding to address them.

Where that growing workload is unpaid, contributors absorb it through their own time, income, and capacity to keep going. Personal commitment cannot supply unlimited hours. If we want safer software while exploitation becomes easier, we must compensate people to spend sustained time reviewing, repairing, and maintaining it. Funding that time is part of the cost of relying on shared software.

Communities already respond through donations, sponsorships, paid maintenance, and other forms of cooperation. That support matters. Yet a beneficiary can have substantial resources and deep dependence on shared work while retaining the option to contribute nothing financially.

The costs remain with contributors, workers, and everyone relying on the resulting infrastructure, while the economic gains can strengthen those who already have the most. Waiting for wider adoption or higher productivity to resolve that imbalance leaves its distribution to the same ownership structures that produced it.

What we believe

We believe substantial resources carry obligations. We believe workers should share meaningfully in the value they create, and the people who sustain shared software deserve a sustainable livelihood.

The responsibility to support others does not depend on whether wealth came from wages, inheritance, ownership, asset appreciation, or another source. It follows from the resources someone holds and the people whose work they benefit from. Reciprocity is part of the social contract that makes cooperation possible.

We believe a license’s ethical value depends on how it serves people. A definition that protects equal access is valuable, but it is not a complete answer to unequal economic power. The freedoms we protect should be judged alongside the livelihoods they sustain and the distribution of benefits they enable.

Software creators should be able to put those convictions into practice. Choosing to share work should not require setting aside beliefs about how its benefits are distributed. The Fair Labor License is our proposed remedy for one part of that problem.

Putting reciprocity into the license

FLL makes source code available to inspect, modify, and redistribute under its terms. It connects free operational use to personal wealth and organizational compensation eligibility.

Under FLL v1.3, individuals with a net worth below US$5 million qualify for free use. Organizations qualify when their CEO’s (or highest-paid executive’s) total annual compensation is no more than fifteen times that of their median employee, subject to the license’s definitions and documentation requirements. Those outside these conditions must obtain a paid license before operational use. Read the full license for the governing terms.

The two conditions apply the expectation of reciprocity to different circumstances. Individuals with substantial wealth are required to contribute financially. Organizations within the compensation-ratio boundary qualify for free use; those outside it must pay the software’s rightsholder.

We chose 15:1 as a practical boundary informed by historical compensation levels and the need for organizations to qualify from the outset. A lower ratio would express a stronger expectation of pay equity, but could exclude too many organizations before the approach had a chance to take hold. The US$5 million threshold makes the responsibility attached to substantial personal wealth concrete.

These boundaries do not certify every labor practice or identify a precise dividing line between fairness and exploitation. A paid license compensates software creators; it does not discharge the buyer’s responsibilities to its employees or guarantee secure maintenance. FLL does not itself redistribute ownership of AI systems or remedy all the inequalities described here. It gives creators a way to require some beneficiaries to contribute financially rather than leave that contribution entirely optional.

That choice introduces eligibility checks and purchasing obligations that unrestricted licenses avoid. It deliberately places FLL outside OSI’s definition; we describe it as source available to make the distinction clear. We believe creators should have this option, and that its obligations should be clear, predictable, and practical to meet.

An invitation to build on it

The communities around shared software have already shown what people can accomplish by building on one another’s work. AI is opening further possibilities for people to create. We want that expanding capacity to come with stronger support for the people behind it and a wider distribution of its rewards.

We invite maintainers, workers, cooperatives, and businesses to help test this approach. Challenge the thresholds. Identify loopholes and unnecessary burdens. Help make the guidance clearer. Share what happens when the license is used, including whether it brings meaningful compensation to the people doing the work.

Different projects will make different choices. We hope FLL contributes to a wider effort to make reciprocity a practical part of sharing software: people can afford to keep contributing, workers share more meaningfully in the value they create, and those with substantial resources help sustain the work they benefit from.

We believe a community that has built so much through shared work can also build stronger ways of supporting the people behind it. We would welcome your help.

Fair Labor License