Responsible AI Policy

Our approach

We see AI as a tool that supports human judgment. We believe people remain responsible for the final decision. Six principles follow from this.

The AI technology we use

MAPRISE builds both generative AI and machine learning into the service itself. AI is not an add-on: it is designed as a core part of how we prepare map data and how you investigate.

Machine-learning image analysis gathers facts across the map; generative AI then explains what they mean. Combining the two, so that grasping the facts and interpreting them form one continuous investigation, is MAPRISE's technical strength.

The team behind the technology

• Our founder, who leads design and development, holds AI, machine-learning and cloud credentials including AWS Certified Machine Learning – Specialty, AWS Certified Machine Learning Engineer – Associate, the JDLA Deep Learning for GENERAL (G Test) and the Generative AI Passport.
• In 2021 the founder won an excellence award at a new-business contest for a diagnostic-support system built on a machine-learning engine for medical images.
• The map data behind AI answers is open data from public bodies that we collect, select, organize and correct. We care about data quality and clear sourcing as much as model performance.
• We have filed two patent applications (on interactive geospatial-information technology, and on technology that analyzes multiple kinds of spatial data together and combines the answers). Both are under examination, and registration is not guaranteed.
• See About the Founder for the full background and credentials.

Machine-learning estimates (such as solar-panel coverage areas) do not guarantee the actual position, shape or area of the facilities.

Principle 1: Support human judgment

• AI answers and analysis are reference information. They do not replace expert opinion or legal judgment.
• They do not replace the investigation and explanation of important matters required under Article 35 of the Real Estate Transaction Business Act, or the research needed for building-confirmation and other legal procedures.
• Always have yourself or a specialist confirm the content before an important decision. A notice to this effect appears in the footer of every page.

Principle 2: Be honest about errors and limits

• AI answers can contain errors (hallucinations), and the same question can produce different answers.
• Map data differs in timing and accuracy by source.
• Sources, terms of use and update status are published in the Data Sources list, and our approach to acquiring data is in the Data Sourcing Policy.
• We provide a way to tell us when you find an error (see "Feedback and improvement" below).

Principle 3: Be fair

• Our Terms of Service prohibit use aimed at unjustly discriminating against, or excluding from transactions, a person or group because of race, nationality, ethnicity, religion, belief, social status, lineage, disability or similar attributes.
• This includes screening counterparties by nationality or religion for tenancy, sale or lending.
• If we find such use, we may take action such as suspending the service.

Principle 4: Protect privacy and information

• We do not use user data, including what you enter into AI chat, for our own purposes such as advertising, AI training or developing new services.
• AI conversations pass through input and output safety checks and filters designed to prevent leaks of personal or internal information.
• AI conversation records carry a signature at the time of storage so tampering can be detected.
• See Our Approach to Security and the Privacy Policy for details.

Principle 5: Be transparent

• The AI chat feature uses Amazon Bedrock (Anthropic Claude).
• We make it clear to users when an answer comes from AI.
• The terms of AI use (credits, prohibited conduct, disclaimers and so on) are in the Terms of Service.
• When we add a new AI-based feature, we will let you know through announcements.

Principle 6: Prevent misuse and keep improving

• The Terms of Service prohibit generating harmful content and prompt injection (feeding the AI malicious instructions).
• They also prohibit bulk data collection by automated means and use for AI training.
• We keep reviewing this policy and our AI mechanisms in light of feedback, internal checks, and developments in technology and law.

Feedback and improvement

If an AI answer is wrong, or wording seems inappropriate, please tell us through the Voice of Customers form. We review it and use it to improve.

Contact: info@maprise.jp

Notice and updates

We may update this policy without notice.