Responsible AI Policy

1. Introduction

AI can create meaningful business value when it’s used with sound human judgment, clear accountability, appropriate security, transparency, data protection, and quality controls. This policy sets out the principles that guide how Binari uses and implements AI, both for our own internal purposes and in the solutions we build for clients.

2. Scope of This Policy

This policy may apply to AI technologies Binari uses or implements, including generative AI, large language models, machine learning systems, AI-assisted development tools, AI-enabled software features, AI Search and answer engines, content generation or analysis tools, AI-powered automation, and recommendation or classification systems.

Specific project requirements may also be governed separately through client agreements, technical specifications, security requirements, DPAs, or sector-specific requirements.

3. Our Responsible AI Principles

We apply ten core principles when using and implementing AI:

1. Human Accountability: final decisions and responsibility stay with people, not with the AI system itself.
2. Purpose and Business Value: AI is used to solve a defined problem, not adopted simply because the technology exists.
3. Privacy and Data Protection: personal data processed through AI follows applicable law and our Privacy Policy.
4. Security and Resilience: AI-specific security risks are considered based on how the system is used.
5. Transparency: we aim to explain AI use when it materially affects a digital experience or business process.
6. Accuracy and Reliability: we acknowledge AI’s limitations and apply reasonable controls to reduce the risk of error.
7. Fairness and Harm Reduction: potential bias and unfair outcomes are considered wherever relevant.
8. Respect for Intellectual Property: AI use takes into account copyright, trademarks, and material owned by others.
9. Responsible Data Use: data processed through AI is used for legitimate, defined purposes.
10. Continuous Evaluation and Improvement: our practices are reviewed periodically as AI technology evolves.

4. Human Accountability and Oversight

Binari treats AI as a tool that supports people and business processes, not a substitute for human accountability. Depending on the risk and context, we may apply human review, approval workflows, manual validation, escalation procedures, or clearly assigned accountability for a given AI output.

The level of human oversight applied reflects the potential impact, the sensitivity of the decision, the data involved, reliability requirements, and any client-specific requirements. We do not claim that every AI output is manually reviewed on every occasion; the level of review is scaled to the level of risk.

5. Purpose-Limited AI Use

AI should be used for defined, legitimate purposes. Before introducing AI into a piece of work, we consider the business problem being solved, whether AI is the right approach, the expected benefits, the likely limitations, operational risk, and whether a simpler alternative already exists. AI is not introduced simply because the technology is available.

6. Privacy and Personal Data

When AI use involves personal data, we seek to minimize unnecessary personal data, use data only for legitimate and defined purposes, follow applicable data protection requirements, and avoid unnecessarily exposing confidential or personal information to third-party AI services. We also assess the role of external AI providers where relevant.

For a fuller account of our data protection obligations, please refer to Binari’s Privacy Policy and DPA.

7. Client Confidentiality and AI Tools

Binari personnel are expected to exercise care before submitting client information into third-party AI systems. Relevant considerations include confidentiality, contractual restrictions, data sensitivity, a provider’s data-use terms, model-training policies, retention settings, and the difference between enterprise and public AI environments.

Confidential client information should not be uploaded to public or uncontrolled AI tools where doing so would violate applicable contractual, privacy, or security obligations.

8. AI Provider and Tool Selection

Binari may evaluate AI providers on factors such as security, privacy, reliability, data handling, contractual terms, model capabilities, where data is processed geographically, enterprise-grade controls, API availability, and operational suitability. The level of formal due diligence applied can vary between providers, depending on risk and context.

9. Accuracy, Hallucinations, and Reliability

We openly acknowledge that generative AI and LLM systems can produce inaccurate information, incomplete answers, hallucinated facts, misread context, or inconsistent results. This is a genuine limitation of the technology as it stands today, not something that can be fully eliminated.

Depending on the application, we may apply controls such as human verification, source validation, structured prompting, retrieval mechanisms, testing, monitoring, confidence thresholds, or fallback processes to reduce that risk.

10. Fairness and Bias

AI systems can reflect limitations in their training data, historical bias, contextual limitations, or choices made in how the model was designed. Where AI affects users or business decisions, we consider the potential for unfair or harmful outcomes and apply appropriate evaluation or mitigation measures. Making an AI system entirely free of bias isn’t a realistic promise; what we can do is keep working to reduce it.

11. Transparency and Explainability

Binari aims to provide appropriate transparency when AI materially affects a digital experience or business process. Depending on context, this can include identifying AI-assisted functionality, explaining the general purpose of an AI feature, disclosing significant limitations, providing a route to human escalation, or documenting key assumptions. Exactly how much explainability is provided can vary based on technical feasibility and the level of risk involved.

12. AI-Generated Content

Where AI assists with writing, summarization, analysis, ideation, code, documentation, SEO content, or AI Search content, the resulting material should be treated as a draft or AI-assisted output rather than a finished product. Depending on its intended use, this may call for factual review, editorial review, originality checks, brand review, legal review, or technical validation, since not all AI-assisted content is written entirely by a human.

13. SEO, AEO, GEO, and AI Search Optimization

Binari’s AI Search Optimization work is intended to help digital content become clearer, more structured, authoritative, machine-readable, useful to real users, and easier for search and AI systems to understand.

Binari does not attempt to manipulate AI systems through deceptive practices, fabricate authority, generate fake citations, impersonate third-party sources, build misleading entities, or deliberately publish false information to influence AI-generated answers.

AI search systems are controlled by third parties, so rankings, citations, recommendations, and inclusion in AI-generated answers cannot be guaranteed. Our methodology also has to keep evolving as search and AI technologies change.

14. Intellectual Property and Copyright

AI use at Binari takes into account copyright, trademarks, licenses, client-owned materials, confidential information, and other third-party intellectual property. Where AI-generated or AI-assisted output is used, we consider the rights attached to its sources, licensing restrictions, originality, attribution requirements, and any contractual ownership arrangements in place. How copyright applies to AI-generated works can vary by jurisdiction and context, so we do not make absolute claims about ownership of such output.

15. AI-Assisted Software Development

AI tools may assist our developers with code suggestions, documentation, testing, debugging, exploring architecture options, or repetitive tasks. AI-generated code is not automatically treated as production-ready.

Depending on the project, appropriate controls can include code review, security review, testing, dependency review, license review, and performance validation. Our engineers remain responsible for production decisions within their assigned scope.

16. Security and AI

AI introduces its own security risks, including prompt injection, data leakage, unauthorized access, malicious inputs, insecure AI-generated code, over-permissioned integrations, and vulnerabilities in third-party models. Appropriate controls are selected based on the system involved and the level of risk. For a broader look at our security practices, please visit our Security & Compliance page.

17. High-Risk or Sensitive Use Cases

Some AI uses call for closer scrutiny because they may involve sensitive personal data, decisions with a significant effect on individuals, regulated sectors, financial or healthcare contexts, employment decisions, safety-critical functions, or legal or compliance decisions.

In these circumstances, Binari assesses the applicable legal requirements, the appropriate level of human oversight, accuracy requirements, data sensitivity, security, potential impact, and any client-specific governance requirements.

18. Prohibited or Restricted AI Uses

Binari does not support, and restricts, AI use for knowingly generating deceptive or fraudulent content, unauthorized impersonation, unlawful surveillance, intentional discrimination, unauthorized use of confidential information, creating malware or malicious systems, manipulation designed to mislead users, deliberately fabricating citations or evidence, or any other AI use that violates applicable law or contractual obligations.

19. Shared Responsibility Between Binari and Clients

Responsible AI is a shared responsibility when Binari develops or integrates AI for a client. Depending on the engagement, the client may be responsible for defining the intended use, providing lawfully obtained data, validating the business requirements, setting approval workflows, establishing sector-specific governance, determining acceptable risk, and ensuring appropriate disclosures to end users. This does not shift all responsibility onto the client; Binari continues to carry its own share as set out in this policy.

20. Testing and Evaluation

AI systems are evaluated in proportion to their use case. Relevant evaluations can include accuracy testing, functional testing, security testing, output quality review, edge-case testing, bias or fairness review, adversarial testing, and performance monitoring. Not every project goes through every form of testing; the scope is scaled to the project’s risk and complexity.

21. Monitoring and Continuous Improvement

AI technology moves quickly. We may periodically review model performance, changes on the provider side, emerging risks, regulatory developments, security issues, business suitability, user feedback, and industry best practice. This policy itself may evolve as well.

22. AI Governance and Internal Responsibility

Responsible AI requires collaboration across relevant functions, including engineering, product, security, privacy, compliance, project management, and business stakeholders. Relevant teams and responsible stakeholders share accountability for applying the principles in this policy within their own areas.

23. Reporting AI Concerns

Clients, users, or other stakeholders with a concern about Binari’s use of AI, such as inaccurate AI-generated content, privacy, potentially harmful output, intellectual property, security, or AI behaving inappropriately, can reach us at:

PT Binary Cipta Solusindo (Binari)
Email: [email protected]

24. Changes to This Policy

This policy may evolve as AI technology changes, laws and regulations develop, industry standards mature, Binari’s services grow, or new risks and best practices emerge. The most recent update date will always appear at the top of this page.

Relationship With Client Contracts

This public Responsible AI Policy expresses Binari’s general governance principles. Project-specific AI obligations may be defined separately through proposals, Statements of Work, service agreements, DPAs, security requirements, or AI-specific project terms. Where a binding client agreement sets out more specific requirements, that agreement governs the engagement.