Protect guesses made about a person's health and beliefs as strictly as information the person shared themselves
Steps to make it happen
- Systems that guess a person's illnesses and views from shopping and search records are spreading.
- Society decides that a guess like "this person seems to be such and such" is also that person's private information.
- To guess a person's illness, religion, or political views, a company needs that person's clear permission.
- People can see what guesses have been made about them.
- People can have wrong guesses fixed or erased.
- Companies can no longer use these guesses on their own to decide who gets a job or a loan.
- A watchdog that belongs to no side checks exceptions, such as a diagnosis at a hospital.
- It becomes normal that guesses about a person are protected as that person's own information.
Water so far 💧 0
The proposal in detail
This proposal covers inferred data, meaning estimates that are worked out by machine from a mix of behavior records, location data, purchase history, and the like, rather than information collected directly from a person. In particular, estimates about things that can easily lead to discrimination or disadvantage if they become known, such as health condition, pregnancy, sexual orientation, religion, political views, ethnic origin, and financial hardship, are placed in the same protected category as the same kind of information when a person gives it themselves. Even when an estimate is not very certain, and even when it is a group label such as "people likely to be ...", it is covered as long as it is tied to a specific person and used. The duty falls on businesses and public bodies that handle data. To make such estimates they need a clear purpose and the person's clear consent, and if the person asks, they disclose that an estimate was made, which items it covered, and how it is used. People can ask for wrong estimates to be corrected or deleted, and using estimates without consent in decisions that greatly shape a life, such as hiring, loans, insurance, and housing contracts, is banned. As exceptions, diagnosis by medical institutions for treatment, statistical research in a form that cannot identify individuals, and public health emergencies are allowed, on the condition that an independent oversight body reviews them.
BenefitsPeople take back control of the inner life they want kept private
The biggest benefit is a check on having illnesses, faith, sexuality, and other things a person has told no one found out and used without permission by combining data. With a system of consent and disclosure, people get a chance to learn what guesses are being made about them and to correct mistakes. This helps reduce hard-to-see discrimination, such as losing a job, a home, or a chance to borrow money because of a label they know nothing about. Minorities and people in hardship are more likely to be hurt by guesses, so the effect of protecting the weak is thought to be large. For businesses too, a clear line on how far guessing is allowed makes it easier to win users' trust, and in the long run it makes services that use data easier for society to accept. Furthermore, by bringing a shared way of thinking to personal data rules that differ from country to country, it can build a base so that, in an age when data flows across borders, inner privacy is protected the same way wherever a person lives.
DrawbacksUseful guesses get blocked, and the costs and the line-drawing remain hard
What opponents weigh most heavily is that tightly restricting guessing itself may stop even uses that help society. For example, a system that spots early signs of depression, or finds elderly people likely to fall victim to financial fraud, may be too late if it has to wait for the person's consent. Also, it is hard to say where a "guess about health or beliefs" begins, and there is a risk that judgments, such as whether showing ads to people who often buy sports gear counts as guessing their health, will vary from business to business. Inside AI, guesses are often not recorded as clear items, so even when asked for disclosure or deletion, there may be no technical way to comply. It is also pointed out that the cost of such compliance weighs more heavily on small businesses and new starters than on large companies that can keep legal and technical staff, which could end up strengthening the position of big companies. Moreover, if the only result is more consent screens and people click "agree" without reading, there is a worry that the protection becomes a mere formality.
Learn the background
In many societies, rules protecting personal information have been built around information collected from the person. But today, by combining data that is ordinary one piece at a time, such as smartphone location, shopping records, and reactions in searches and on social media, it has become possible to guess with fair accuracy things a person has told no one, such as illness, pregnancy, faith, political views, and sexuality. With the development of AI, these guesses are made more cheaply and in far greater volume. The problem is that information made by guessing easily falls outside what must be protected, or its handling becomes unclear, because it was "not provided by the person." People may be treated unfairly, not only in which ads they see but in hiring, loans, and insurance decisions, without knowing how they have been guessed about or what category they have been put in. If a guess is wrong, a label they know nothing about spreads without their knowledge. Now that the power to read a person's inner life from data has grown stronger, the thing to protect needs to be widened from "what was collected" to "what became known."
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 2026-10-15d) and translated by AI (claude-sonnet-5-5). Nobody has checked the facts. Read the original (日本語) How the AI writes