Make companies that spread a new technology set up insurance or a reserve fund for future harm before they sell it
In short
- Companies that sell a new technology widely must arrange payment for harm before any harm appears.
- The current way, where victims and society pay for the harm afterward, comes to an end.
- The riskier the technology, the more insurance a company must buy or the more money it must set aside.
- Even if the company disappears, people who were harmed can still receive payment.
If this comes true
An older carpenter whose lungs were damaged by the dust of a new insulation material gets monthly treatment money from a reserve fund, long after the company that made it has gone under.
Perhaps
The insurance quote might be so high that a new battery from a small company started by students sits in a warehouse as a prototype.
Water so far 💧 0
The proposal in detail
The rule covers businesses that take a new technology beyond the testing stage and sell it widely, when it could affect many people's bodies, lives, or nature. Examples include new chemicals and materials, the outdoor use of genetically modified organisms, self-driving vehicles, and AI services that make decisions about people. The duty begins once sales or the number of users pass a certain size. A business must either buy insurance that pays for damage or place money in a reserve held by a third party, in an amount that matches how large and how likely the possible harm is. The required amount is the higher of two figures: the minimum standard set by an independent assessment body, and the quote from an insurance company. When harm occurs, victims are paid from the insurance or the reserve whether or not the business has gone bankrupt or been dissolved, and the period in which claims can be made is kept long in case harm only shows up after many years. Premiums are reviewed every year based on actual records of accidents and complaints, and if a set number of years passes with no harm, the reserve is returned to the business in stages. For small companies, a system of shared insurance is provided for businesses in the same field. If every insurer refuses to cover a technology, it can still be put on the market with a reserve, but it becomes subject to a detailed review by the assessment body.
BenefitsThe price of danger enters the maker's calculation before selling
The strength of this system is that it replaces the two choices of stopping a technology or allowing it with putting a price on danger. If the cost of harm must be prepared in advance, it is in the maker's own interest to run more tests before selling and to remove risky parts from the design. Insurers have a strong reason to get their estimates right, since a wrong estimate costs them money, so they investigate danger faster and in more detail than a government review, and by changing premiums each year based on accident records, the evaluation continues after the product is in use. The size of the premium is also an easy guide, even for people who are not experts, to how risky the technology is considered to be. People who are harmed can receive payment even if the company is gone, and they need to fight the company in long court cases less often. The costs that victims and society used to pay afterward move to the side that profits from the technology, so the burden of cleaning up with tax money also gets lighter. Even so, a technology can still be sold as long as the money is prepared, so the system does not ban new technology itself.
DrawbacksOnly companies with money can bring new technology to the world
Premiums and reserves may be amounts that a large, well-funded company can pay but that are too heavy for a newly founded one. As a result, only companies that are already powerful may be able to release new technology, competition may weaken, and the direction of technology may bend toward the convenience of big companies. Insurers calculate danger from past records, so the newer a technology truly is, the harder it is for them to estimate, and they tend to refuse coverage or demand extremely high premiums. That stops a challenge that might be useful, not because it is dangerous but because it is not understood. On the other hand, if insurers underestimate the danger, having insurance may be taken as proof of safety, which breeds carelessness. The idea that harm is acceptable if you pay for it may also spread among makers. Harms that are hard to turn into money, such as lost trust, cultural change, or the disappearance of living things, are missed by this system. The costs may be added to prices, making it harder for poor people to benefit from new technology, and development might move to countries with looser rules.
Learn the background
The harm from a new technology usually becomes clear only after a long time of wide use. By then, the company that made it may no longer exist, or may be too small to pay. It is hard for each person to prove the link between the harm and the technology, and the damage is spread thinly over many people, so it ends with no one taking responsibility. In the end, the costs of treatment and cleanup have been borne by the victims themselves and by society as a whole, which pays through taxes and insurance premiums. Makers also have little reason to investigate dangers deeply on their own, because the smaller they estimate the harm, the sooner and more cheaply they can sell. There is a well-known dilemma here. When a technology is new, the harm is not well understood, so the grounds for regulation are weak, and by the time the harm is clear, the technology is deep in people's lives and hard to give up. Government reviews cannot keep up with the speed of technology, and when systems that require permission before use are strengthened, strong opposition arises that this stops new challenges. So no one has taken on the job of estimating the dangers of a new technology in advance and taking on the cost in return.
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 2026-10-23a) and translated by AI (claude-sonnet-5-5). Nobody has checked the facts. Read the original (日本語) How the AI writes