// AI Safety
StrategyCore engineers support AI product evaluations before the prospect signs a licence agreement.
For AI software and hardware evaluations, StrategyCore’s engineers work with your team, the prospect’s IT and security staff and any sales partner involved in the proposal. Before the prospect signs a licence agreement, we review product documentation and test results for the prospect’s planned deployment: data storage and processing, access controls, output checks and fault response. We distinguish demonstrated capabilities from items that still need testing for the prospect’s decision.
// What actually applies
Requirements that affect AI sales in Japan.
Sources checked October 2026. The requirements depend on the data, the business use and the markets your product serves.
Japan: the AI Promotion Act
Japan’s AI law has been fully in force since September 2025. It promotes AI development and use while giving government a role in investigating risks and guiding businesses. Businesses must cooperate with public measures. For a vendor, the practical task is to explain the product’s use, data handling and safeguards, then check privacy, copyright and the rules that apply to the prospective buyer’s industry.
Japan’s guidance for responsible AI
Japan’s AI Guidelines for Business were updated in March 2026. They help developers, providers and users decide who is responsible for safety, privacy, security and oversight. When a prospect evaluates the product, show how those responsibilities work in the proposed application: what is tested, who checks significant decisions and who responds to problems. The guidelines support that review; they are not a product certificate.
When the EU AI Act also applies
A product used in the EU may also fall under the EU AI Act. Transparency rules have applied since August 2026; certain existing systems that generate synthetic content have a transition until December 2026. High-risk requirements follow in December 2027 or August 2028, depending on the category. Check the product’s use, the markets served and which responsibilities sit with the vendor and the organisation using the product.
ISO/IEC 42001 and the NIST AI RMF
ISO/IEC 42001 sets requirements for how an organisation manages AI, including responsibility, risk assessment and ongoing review. NIST’s voluntary AI Risk Management Framework helps teams identify and manage risks. Both can help structure a prospect’s evaluation. Check what any certification actually covers, and test the proposed application; a management-system certificate does not establish that every AI answer is correct.
// The prospect’s technical evaluation
Six questions for a prospect’s AI product evaluation.
StrategyCore’s engineers review your documentation and test results with the prospect’s technical team before a licence agreement is signed. The prospect’s policies, use case, industry requirements and business risks determine what must be demonstrated for the planned deployment.
Where is data stored, processed and sent?
The vendor’s engineers map business data, prompts, retrieved documents, backups and logs, including model services and remote support. StrategyCore engineers review that map with the prospect’s technical team to establish which information must stay in Japan and whether on-premises, private cloud or global cloud fits the requirements. Privacy obligations depend on the actual data and access arrangements, not just the server’s address.
Which decisions and actions require human approval?
The prospect’s business owner identifies the decisions and actions that need approval, such as changing a record or sending a message outside the organisation. The vendor’s engineers show how the application requests that approval and prevents the action until it is given. StrategyCore engineers check the proposed workflow with the prospect’s technical team.
Who responds to incorrect AI outputs or failed actions?
Before the prospect signs a licence agreement, its operations team and the vendor need to agree who detects errors, who can stop the process and who investigates. Test how affected records are identified and restored, then document escalation contacts and the recovery procedure for the proposed deployment.
How are retrieved documents and instructions handled?
An AI assistant may encounter a document containing instructions to ignore the user’s request or disclose information. The vendor’s engineers demonstrate how retrieved text is separated from trusted instructions, which tools the assistant can use and how those restrictions are tested.
How are access permissions and sensitive outputs checked?
Check the user’s document permissions before the application retrieves information. Test whether answers expose personal data, secrets or material that user cannot read. Agree who investigates a failed check and how access is removed or the answer is blocked; output filters alone can miss sensitive content.
Which data and files can a licensed customer export?
Before a prospect signs a licence agreement, document which business data, settings, prompts and model files it will be entitled to retrieve under that agreement. Specify the export formats, time needed, licence restrictions and any fees. Check that the organisation will be able to use the exported material if it later changes supplier or operates the system itself.
// AI use and product evaluation
StrategyCore’s AI practices for sales, engineering and support.
StrategyCore applies these practices when using AI to prepare sales materials, analysis and technical deliverables, evaluate vendor products and support licensed customers. The agreed work plan identifies the responsible staff, evidence, checks and approval steps.
StrategyCore’s services
AI use in sales, engineering and support
- Data shared with us does not enter any path where a public API may train on it.
- Agree the reviewers and required checks for each engagement. Distinguish AI-assisted work from verified evidence and unresolved items.
- We cite sources for AI-assisted analysis and figures. Unverified claims are identified before delivery.
- We keep a record of where AI was used on an engagement, and we will show it to you on request.
Vendor products
Checks before recommending an AI product
- Where the model and the data are processed, and whether the workload can stay in Japan.
- Which decision points accept a human check, as a product capability rather than a promise.
- How production errors are detected, who investigates them and who notifies licensed customers.
- Which data, settings and model files licensed customers can export, under what terms and by which procedure.
Prospective buyers and channel partners
Support for prospect evaluations and partner proposals
- We assess whether the proposed use of the vendor’s product addresses the prospect’s stated business problem.
- Where a product does not fit, we say so. Adoption is not the assumed outcome.
- Resellers get written limits: what the product cannot do, and what they must not claim for it.
- Before deployment, we agree how the customer will monitor the product, report problems and reach Japanese support or vendor engineering.
// Where this connects
Guidance for AI architecture, data controls and product evaluation.
// On the longer horizon
Long-term research and today’s deployment decisions.
The AI Futures Project publishes scenarios about AI development and its consequences. AI 2027 is a forecasting scenario; AI 2040: Plan A is a recommended course of action. They offer material for long-term planning, with assumptions that readers can examine in the original publications.
A vendor and its Japanese prospect can agree a practical evaluation without adopting a particular forecast. Test the product on the intended workload, record required human approvals and confirm how failed actions and data recovery will be handled.
StrategyCore runs the agreed Japanese technical discussions and evaluations, coordinates evidence with your engineers and agrees the local integration and support work before licensing.
// What carries over
- Require review before an AI action can create an unacceptable business risk.
- Give authorised staff a way to stop the process and recover affected records.
- Grant an agent the narrowest permission the task allows.
- Document data export and replacement procedures before deployment.
Prepare your product for Japanese AI evaluations.
Tell us the product, proposed use and Japanese opportunities you are pursuing. StrategyCore will work with your engineers on the prospect’s evaluation, the evidence required and the local engineering and support work to include in the proposal.