AI Poisoning Prevention System

As AI using machine learning and deep learning has begun to spread rapidly in recent years, there is a risk that the input data to AI can become contaminated (data poisoning), leading to AI malfunction.

For this reason, it is essential to guarantee the safety and reliability of input data, but currently there are no systems that do so.

In the case of data input/output with AI via the cloud, the security of input and output data must be ensured and protected in the same way.

Furthermore, there is a risk that AI products may infringe on copyrights because the generated AI does not prove that the learned data does not contain copyrights or other intellectual property rights, or that it does not contain other pre-existing AI products.

In addition, there will be more and more applications that require proof that the training data and output data from AI are politically correct (neutral expressions that do not include prejudice or discrimination based on race, religion, gender, or other differences).
As a result, the environment is not yet conducive to the safe and maximum use of AI.

DATA POISONING

Risk of AI dysfunction and malfunction due to contamination of input data to AI (data poisoning)

SECURITY RISK

Security risks when communicating data input/output with AI via cloud

COPYRIGHT INFRINGEMENT

Risk of litigation for copyright infringement by its AI products

POLITICAL CORRECTNESS

Risk of fire and litigation related to political correctness

Prevention of input data contamination
Data Poisoning Prevention

Data used for training and decision making to AI can be recorded on the blockchain and data can be retrieved via the blockchain to prevent tampering.

Input/output data tracking

Recording input/output data to/from the AI on the blockchain prevents tampering and enables tracking of where the input came from and where it was sent to, etc.

Proof of data entry history

The history of which data was input and by whom can be recorded in the blockchain to prove the history of data input to the AI. These can prove that the AI is being trained only with original data, and can provide evidence in the event of a lawsuit.

Validation of input data

Data legitimacy is enabled by the AI front-end functionality, which assigns legitimacy information for input and output data with the AI and records it in the blockchain to prove that it is not incorporating unauthorized data that disregards copyright or data that is tainted or amounts to political correctness.

In the process of training AI and integrating it into products and production lines, it is critical that the AI obtain correct output data without contamination.
Also, in the event of a malfunction in the operation of AI after it has been shipped as a product, the legitimacy of the training data can be proven, thereby avoiding the risk of litigation.

Drug discovery and food-related products are being created based on know-how-based technologies, and attempts are being made to improve development efficiency through the introduction of AI.
However, once fraudulent or contaminated data are introduced into the AI that has been continuously trained, finding and fixing them is not an easy task.
Keeping AI safe from contamination and continuing to prove the validity of data will allow for continued safe development.

There are increasing opportunities to use generative AI in the creative field.
In such cases, the training data can be recorded in the blockchain to ensure that the AI does not make expressions that amount to copyright infringement or political correctness, thereby proving the legitimacy of the AI product.