"Private AI" text anonymization technology to promote the use of generative AI and LLM

DX business menu

DX business
HOME

Solution

Handling Manufacturer

event·
seminar

Document·
Case studies and columns

Inquiry

TOP

Seminar content

seminar

Document request

inquiry

Private AI

Data privacy hinders the use of AI

As the use of generative AI in business operations rapidly permeates companies and society, the risk of AI being misused is increasing dramatically. In particular, there are a wide range of considerations under the Personal Information Protection Act, and careful consideration is required.

<Challenges when dealing with AI risks such as personal information protection>

Actual unique data is difficult to convert into masking data

Building privacy solutions in-house takes a lot of time and effort

Running self-trained transformer models takes time and leads to huge cloud charges

Personal information protection laws are constantly evolving, and it is extremely important to stay on top of new requirements.

Text anonymization technology “Private AI” to promote the use of generative AI/LLM

Private AI helps users input information into large-scale language models (LLMs) and fine-tune models using their own data. A service that uses unique machine learning algorithms to automatically edit confidential information and personally identifiable information (hereinafter referred to as PII), enabling the use of AI while ensuring data privacy. We offer

4 features

Accuracy

Overwhelming accuracy

Wide range of coverage

Supports PII, PCI, PHI in 52 languages and 50+ types

Privacy and security

Deploy via containers on-premises or in private/public clouds, so your data never leaves your environment

compliance

the Company containers are APPI, GDPR, CPRA and HIPAA compliant

1 Accuracy

PII detection using AI trained on large amounts of data, rather than rule-based, enables highly accurate detection.

2 Wide range of coverage

Supports 52 languages including Japanese and over 50 entities.

Personal information PII
- full name
- 年齢
- Address/place name
- date
- Birthday
- Driving license no.
- My Number (Social Security Numbers)
- phone number
- URLs

Also supports entities unique to Japan
- My number
- Japanese address
- Japanese phone number

Personal Health Information PHI
- blood type
- Condition (medical condition)
- Drug name
- drug dosage
- physical injury
- Treatment process (treatment, procedure, test name, etc.)

Credit information PCI
- Bank account information
- Credit card no.
- credit card expiration date
- CVV
- Routing Number
- Bank code
- Branch number etc.

For other entities, please feel free to contact us.

3 Privacy and security

Deployed via containers on-premises or in private/public clouds, your data never leaves your environment and remains secure.

4 Compliance

Containers are APPI, GDPR, CPRA and HIPAA compliant

Use Case

Can be used in various situations in AI applications where data privacy is required.

Finance/Insurance

Credit screening
Fraudulent payment detection

medical care

Diagnosis automation

retail

Fraudulent payment detection
marketing

Use case 1: Anonymization of analysis data

Since the data is personal information, it is strictly managed.
High hurdles to utilizing data analysis

Challenges in anonymization
- It takes a lot of man-hours to process masking manually or using regular expressions.
- Unstructured data is difficult to process automatically
- I don't know what information I should pay attention to in terms of GDPR etc.
- Rather than anonymizing all data, I want to preserve the data necessary for analysis, so I want to change the anonymization target for each analysis theme.

When handling PII such as names and addresses, data can be analyzed safely by anonymizing unnecessary information.

Use case 2: LLM collaboration

Personal information may be included in LLM prompts
If the LLM service is an external service, personal information will be leaked outside the company.

Automatically block personal information and prevent leakage by filtering Pronto