Proposed Corrective Measures for Generative AI (2025/11/02)

This time, I will share my proposed corrective measures for generative AI.

I believe there are three major problems with generative AI.

1. Opaque training methods

It uses opaque training methods where the source of the training data is not disclosed.

Even if the training itself is legal, if the source material is in a gray area, one cannot use it with peace of mind.

2. Exploitation of training sources

The fact that it exploits training sources without permission is also a problem.

It is painful for the users as well.

3. Risk of accidental copyright infringement

The current reality of generative AI is that we cannot deny the possibility that generated content might accidentally infringe on copyrights.

Unless this is improved, we cannot use the generated output with peace of mind.

Based on the above issues, I propose the following corrective measures.

1. Cleaning up the training process

Disclose the training data to ensure transparency.

2. Returning compensation to training sources

Pay compensation to the training sources and do not train without permission.

3. Copyright check function

Implement a function that automatically checks whether generated content infringes on copyrights.

These are the corrective measures for generative AI that I have considered.