Let makers pay for research that tests their technology's safety, but not choose what the research looks at
In short
- Companies pay the cost of safety research on their own products into a shared fund.
- An independent fund, not the company, decides what gets studied and which researchers do it.
- Companies must hand over the data and samples that researchers need for their work.
- All results are made public, whether they help the company or hurt it.
- A company can no longer be approved as safe on the strength of studies it chose itself.
If this comes true
A young researcher who studies sleep publishes the finding that a popular video app cuts into children's sleep, and the company that paid for the work never saw it first.
Perhaps
While a small company waits for the fund's review, a useful machine it built might stay off the market for years, and its makers might have to close their shop.
Water so far 💧 0
The proposal in detail
The rule covers products and services whose research results are used for government approval or for claims about safety and effect: medicines, farm chemicals, food additives, medical devices, AI systems, and apps and social media with many users. When makers show safety or effectiveness in a regulatory review or in advertising, they pay for the research behind that claim into a public research fund that is independent of both the government and the industry. The amount depends on the size of their sales and on how serious the risk to be checked is. The fund's board includes no one who is paid by the industry in question, is made up of researchers, doctors, user representatives and others, and makes public the members' terms and any conflicts of interest. The fund registers and publishes the questions and methods in advance, chooses researchers by open call, and publishes the full results whether they are good or bad. Makers are obliged to provide the samples, data and access to AI models that the research needs, under confidentiality agreements, and if they refuse, approval or continued sales are not allowed. Makers' own in-house tests are not banned, but they cannot be the only basis for approval. A company may fund a university study of the harms of its own product directly only by going through the fund. Basic joint research and research unrelated to product safety are outside the rule. Companies with small sales get a lighter burden, and in urgent situations such as an outbreak of infectious disease, independent research runs alongside a provisional approval. Countries recognize one another's fund results and share them so that countries without their own research systems can use them too.
BenefitsEvidence is freed from the seller's interests, and harm is found sooner
When the power to choose research questions leaves the company, the least visible bias disappears: inconvenient questions never being asked in the first place. Researchers do not have to worry that their funding will be cut for publishing an unfavorable result, so they can publish freely, and early signs of harm reach society sooner. Because access to data becomes an obligation, the effects of social media and AI, which have been closed inside companies, can be examined by outside eyes. Regulators can obtain trustworthy evidence without a big budget of their own, and the exhaustion of regulators and industry accusing each other of bad faith with every decision is reduced. The cost is paid by the makers who profit, so taxes do not have to rise much. In addition, for companies that honestly make safe products, the fact that independent research confirmed them becomes a source of trust in itself, and they are no longer treated the same as rivals who get by with convenient studies. Because the same system can be used in any field, the work of writing new rules from scratch each time a new technology appears also shrinks.
DrawbacksWith one road to verification, the risks of delay and bias pile up in one place
Gathering the power to decide content into one fund also gathers the damage when that fund is wrong or captured. If the influence of the government or of a particular school of thought enters the choice of board members, a new bias could arise that treats one technology unfairly harshly and another leniently, and companies would find it hard to prepare and present their own research to answer it. Reviews and open calls take time, the period before approval gets longer, and useful medicines and machines reach patients and users later. Because the lives and opportunities lost to that delay are hard to see, they tend to be treated lightly. The fee is added to product prices, small and new companies feel it most heavily, and there is a risk that only large companies remain in the market. Handing data and models to outsiders also risks leaking trade secrets and spreading clues that could be misused. And if development bases move to countries that do not adopt the system, a reversal could occur in which the countries that were strictest are the slowest to get new technology.
Learn the background
Much of the evidence that decides whether a new medicine, farm chemical, AI system or app is safe comes from research paid for by the very companies that profit from selling it, with the questions and methods chosen by them. Looking back, it has been pointed out again and again that for tobacco, some foods and chemicals, industry-funded research made harm look smaller and delayed regulation by decades. Today, even when outside researchers want to study how social media affects young people's minds or what dangers AI poses, the key data is in the companies' hands and they cannot get near it. The people in trouble are those who use the products, researchers who want to report honest results, and regulators who are pressed to make decisions. Past measures centered on disclosing conflicts of interest, which means stating in a paper where the money came from. But disclosure only shows who paid. It does not change the structure in which the payer chooses the questions. Public research money is short, universities depend on corporate funds, and companies want to avoid unfavorable results. Regulators have no budget to investigate on their own and have to use the materials the companies provide. As long as the person who pays is also the person who decides the content, this bind cannot be undone.
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