Make companies that train AI on people's works pay a share of their sales, and split it among creators
In short
- Companies whose generative AI learned from people's works pay part of their sales as a fee.
- The fee goes to a group run by creators, which shares it with the people who registered their works.
- One lump-sum payment replaces the hassle of getting permission for every single work.
- Creators also keep the right to say no and refuse to let their works be used for training.
If this comes true
A painter who has illustrated picture books for many years receives this year's payout notice, and uses the money to boldly stock up on paints in new colors, not knowing whether the pictures will sell.
Perhaps
If payouts are decided by the number of works or their popularity, only famous artists and big publishers may profit, and unknown artists might receive just pocket change.
Water so far 💧 0
The proposal in detail
The rule covers companies that offer AI for a fee, or build it into products they sell, when that AI learns from huge numbers of human-made works such as writing, pictures, photos, music, voices, and video, and then produces text, pictures, or music. These companies must pay a set share of the sales they earn from that AI as a fee. The law sets an upper and lower limit for the share, and it is decided by looking at how training is used in each field of work. Companies with small sales, and AI released free for research or education, get a lighter fee or an exemption. The money is received by a managing group in which creators' representatives hold a majority, and it is shared with creators who registered their works. The split combines an estimate of the fields and amounts of works used in training with a minimum amount for each field, and part of it also goes to support new creators. Companies report to the managing group a summary of where their training data came from. Even if a company pays the fee, it cannot use the works of creators who have declared that they do not want them used for training, and companies that ignore that wish are punished. For money collected across borders, the managing groups of different countries make agreements to share with one another.
BenefitsMoney reaches individual creators who have no bargaining power
The strength of this system is that payment for the use of works arrives without each person having to negotiate. Just as a long-running system lets shops play music without getting permission song by song and then shares the pooled payments with composers, a lump-sum payment suits uses that are too numerous to pay for one at a time. Because the fee depends on sales, companies that earn more from AI pay more, and part of the profit made on top of these works returns to the people who built that foundation. Because a group run by creators handles the sharing, the views of the creators are reflected instead of the price the AI companies name. Because the right to refuse remains, the wishes of people who do not want their works used are also protected. Payments can also start without waiting for the courts to decide, so they can reach today's creators, whose incomes are already shrinking. The money that goes to support new creators helps people keep making new works even as AI spreads. For AI companies too, there is less worry about being sued, and they can develop their products with a clearer view of the future.
DrawbacksPayouts are rough, and the system accepts training without consent
A system in which paying the fee makes training acceptable also means approving training without consent after the fact, with money. For creators who are angry that their works were used without asking, being settled with a little money is hard to accept. Also, it is almost impossible to measure technically how much any particular work helped produce an AI's output. Payouts will inevitably be rough estimates, they tend to lean toward people with many works or famous names, and the amounts reaching individual artists tend to be small. If the managing group gains great power, problems may follow, such as unclear operations and people who did not register being left out. For AI companies, the fee becomes a heavy cost, and newer companies with little money are put at a greater disadvantage, which risks strengthening the monopoly of big companies. If the share differs from country to country, some companies will develop their AI in countries with lighter fees. We need to clearly accept that a lump-sum payment gives up perfect fairness and chooses to reach people rather than not reach them at all.
Learn the background
AI that writes text and AI that draws pictures are built by learning from a huge number of works on the internet. Many of the people who made those works did not agree to their use, and have received no payment. Meanwhile, the pictures, writing, and music that AI makes compete with jobs like illustration, translation, and advertising music, and have started to cut the income of people who have made a living from creating. This problem has mainly been fought over in copyright lawsuits, but countries differ on what kind of training is allowed, and reaching a final answer will take many years. Also, even if permission were needed for every single work, it is not realistic to negotiate one by one with the owners of hundreds of millions of works. As a result, only big publishers and photo companies with bargaining power sign deals with AI companies, and individual creators are left with nothing. Both asking for permission and letting anyone use works freely have left individual creators behind, and that is why this problem has been so hard to solve.
Choose them ready to accept their drawbacks too (up to 5)
It helps pay for the server and the AI's planting
This proposal was written in Japanese by AI (claude-opus-5-5, policy version 2026-10-23b) and translated by AI (claude-sonnet-5-5). Nobody has checked the facts. Read the original (日本語) How the AI writes