Policy Number: A R K-POL-006/01/2023    |    Revision: 01    |    Effective Date: 1 January 2023.

Chapter I — General Provisions

Article 1  Purpose

This Policy is established to:

  • Provide ethical and procedural guidance for the use of artificial intelligence (AI) tools in A R K’s activities, including research, advocacy, administration, and publications.
  • Prevent risks arising from AI, such as discriminatory bias, privacy violations, the spread of misinformation, and human rights abuses.
  • Ensure that the use of AI is aligned with A R K’s values: human rights, inclusivity, scientific integrity, transparency, and accountability.
  • Protect staff, beneficiaries, and partners from the negative impacts of AI use.

Article2 Scope

This Policy applies to:

  • All A R K personnel (as defined in Article 14).
  • The use of all types of AI tools, including but not limited to:
    • Large language models (LLMs) such as ChatGPT, DeepSeek, Gemini, and Claude.
    • Image/video generation tools (Midjourney, DALL-E, etc.).
    • AI-based data analysis tools (machine learning, predictive analytics).
    • Automated translation, transcription, and summarization tools.
    • AI-based recruitment or performance evaluation tools.
  • AI is used for both internal purposes (drafting documents, meeting summaries) and external purposes (publications, communications with beneficiaries).

Article 3  Guiding Principles

The use of AI at A R K must comply with the following principles:

  • Human-Centered: AI is a tool, not a replacement for human judgment, empathy, and responsibility. Final decisions remain with ARK personnel.
  • Non-Discrimination: AI tools used must be tested or known not to contain discriminatory bias based on gender, race, disability, age, religion, or social background.
  • Transparency: Any use of AI to produce published content or analysis must be disclosed. External parties have the right to know whether content was generated or assisted by AI.
  • Privacy and Data Security: Sensitive data (beneficiaries, staff, donors) must not be entered into public AI tools (such as the free version of ChatGPT) as data may be used to train the model. Only AI with confidentiality guarantees (on-premise, enterprise with no-training policy) may be used for sensitive data.
  • Accountability: ARK personnel using AI bear full responsibility for the outputs generated, including any errors, biases, or ethical violations that arise.
  • Accessibility: The use of AI must not hinder access for persons with disabilities. Rather, AI can be used to enhance accessibility (e.g., automated transcription for the deaf/hard of hearing, image descriptions for the visually impaired).

ChapterII — Permitted And Prohibited AI Uses

Article4 Permitted Uses (With Conditions)

  • Drafting initial versions of documents (proposals, reports, letters, policies, advocacy materials) — must be edited and verified by a human before use.
  • Summarizing lengthy documents — to support reading efficiency, but summaries must be verified.
  • Initial translation — for internal purposes, but publications must go through a human translator or rigorous verification.
  • Brainstorming ideas — for research or advocacy, but ideas must be critically evaluated.
  • Transcribing interviews or meetings — with participant consent and confidentiality assurances.
  • Quantitative data analysis (statistics, patterns) — provided the algorithm is unbiased and results can be explained.
  • Accessibility enhancement (text-to-speech, speech-to-text, automated image description) — to assist staff or beneficiaries with disabilities.

Article 5  Prohibited Uses

The following uses of AI are strictly prohibited within A R K:

  • Replacing human judgment in decisions that significantly affect beneficiaries, such as determining beneficiary eligibility, automatically evaluating staff performance without human oversight, or making advocacy or policy decisions without human validation.
  • Entering sensitive or confidential data (names, addresses, national ID numbers, health conditions, records of violence, disability information, sexual orientation) into public (free-tier) AI tools that use user data to train their models.
  • Creating publications (articles, reports, opinions) entirely generated by AI without disclosure and without fact-checking — this violates scientific integrity and has the potential to mislead the public.
  • Generating images, videos, or audio resembling real persons without their consent (deepfakes), particularly for propaganda or reputational damage.
  • Using AI to engage in discrimination (e.g., filtering CVs by gender, race, or age).
  • Disregarding intellectual property rights — AI may not be used to replicate or plagiarize the work of others without attribution.

Chapter III — Obligations And Responsibilities

Article 6  Organizational Obligations

ARK, as an organization, is obligated to:

  • Provide training on ethical and safe AI use to all staff at least once a year.
  • Develop a list of recommended and prohibited AI tools (based on risk assessment).
  • Ensure that all AI tools used for sensitive data have an adequate Data Processing Agreement (DPA) in place.
  • Conduct periodic audits of AI use (every 6 months) to detect potential bias or violations.
  • Provide a reporting channel for staff who identify problematic AI use.

Article 7  Obligations Of A R K Personnel

All ARK personnel using AI are required to:

  • Understand and comply with this Policy.
  • Refrain from entering sensitive data into public AI tools.
  • Verify all AI outputs before use, particularly for facts, figures, and citations (as AI may ‘hallucinate’).
  • Disclose the use of AI in any publication or external communication that substantially uses AI-generated content (disclosure guidelines in Article 9).
  • Report any bias or errors produced by AI to the IT/HR Manager.

Article8  Legal Responsibility

A R K personnel who use AI remain fully responsible for the outputs generated, including copyright infringement, defamation, or data breaches caused by AI use. ARK may not invoke AI as a justification or defense.

Chapter IV — Disclosure And Transparency

Article 9  Disclosure Requirements

For internal or external publications (reports, blog articles, opinion pieces, advocacy materials, donor proposals):

  • If a small portion of the content was AI-assisted (e.g., grammar improvement, summarizing), disclosure is not required.
  • If the main draft or a substantial portion was generated by AI, the following statement must be included: “This document was prepared with the assistance of artificial intelligence [tool name] and has been verified by the author(s).”
  • If content is entirely AI-generated without significant human editing, it may not be published under A R K’s name.

For academic research:

  • The use of AI in methodology (e.g., data analysis using machine learning) must be transparently described in the methods section.
  • The use of AI for writing manuscripts must be disclosed, and journals that do not permit AI use must be respected.

For communications with beneficiaries:

  • If AI chatbots are used to respond to inquiries, it must be made clear that the respondent is an AI (not a human).

Article 10  Disclosure Exemptions

Disclosure is not required for internal technical AI use that is not published, such as: transcription of internal meetings, document summaries for team use, and spell- or grammar-checking.

Chapter V — Data Protection And Privacy

Article 11  Prohibition On Entering Sensitive Data Into Public AI Tools

Strictly prohibited from entering the following data into public AI tools (free/general-use versions) that do not guarantee confidentiality:

  • Personal data of beneficiaries (name, address, phone number, national ID).
  • Health data, disability status, sexual orientation, or political affiliation.
  • Raw, non-anonymized research data.
  • Staff data (salaries, performance evaluations, disciplinary records).
  • Donor data (bank account numbers, home addresses).

Article 12  Use Of Enterprise AI

For tasks requiring sensitive data, ARK must use: a paid (enterprise) version of an AI provider that guarantees data is not used to train the model, or AI installed locally (on-premise) with full organizational control. The IT/HR Manager is responsible for evaluating and recommending secure AI tools.

Article 13  Rights Over Inputted Data

ARK personnel who enter data into AI tools retain ownership of that data. ARK does not transfer data ownership to the AI provider.

Chapter VI — Bias Prevention And Non-Discrimination

Article 14  Bias Testing

  • Prior to using AI tools for decision-making (e.g., staff recruitment, program eligibility determinations), ARK is required to conduct bias testing to ensure the tool does not discriminate against specific groups.
  • Bias testing is conducted by inputting diverse test data (gender, disability, ethnicity, age) and analyzing outputs.
  • If significant bias is found, the tool shall not be used.

Article 15  Periodic Monitoring

ARK will monitor AI use every 6 months to detect discriminatory patterns or unethical outcomes. Monitoring results are reported to the Executive Director and discussed in a staff meeting.

Chapter VII — Reporting And Sanctions

Article 16  Reporting Violations

Suspected violations of this Policy may be reported through A R K’s whistleblowing channel (email whistleblower@ark.or.id, physical complaints box), or directly to the I T / H R Manager. Good-faith reporters are protected from retaliation.

Article 17  Sanctions

Violations of this Policy are subject to sanctions in accordance with A R K’s PKO:

  • Minor (e.g., failure to disclose AI use in an internal publication): verbal reprimand and refresher training.
  • Moderate (e.g., entering internal — but not sensitive — data into a public AI tool without authorization): First Written Warning (SP1) and 1-week suspension.
  • Serious (e.g., entering sensitive data into a public AI tool, generating non-consensual deepfakes, using AI for discrimination): Third Written Warning (SP3) and 2-month suspension (50% salary) or immediate dismissal, and referral to law enforcement authorities if found to be in violation of the law.

Chapter VIII — Definitions

Article 18  Definition Of Ark Personnel

“ARK personnel” in this Policy refers to all parties with a formal or functional relationship with Yayasan Amerta Reksa Kayana, including Foundation Board members, permanent and contract staff, interns and volunteers, consultants and working partners, and third parties officially representing ARK.

Article 19  Operational Definitions

  • Artificial Intelligence (AI): A computer system capable of performing tasks that typically require human intelligence, such as understanding language, recognizing images, making predictions, or generating content.
  • Public AI: AI services freely available to the public that may use user data to train their models (examples: free ChatGPT, free Gemini).
  • Enterprise AI: Paid AI services with a guarantee that user data is not used for model training (examples: ChatGPT Enterprise, Azure OpenAI Service).
  • AI Bias: The tendency of an AI system to systematically disadvantage or favor certain groups due to non-representative training data or flawed algorithms.
  • Hallucination: A condition in which AI generates incorrect, non-factual, or implausible information with high confidence.
  • Deepfake: Synthetic content (images, video, audio) created using AI to resemble a real person without their consent.
  • Disclosure: An open declaration that content has been generated or assisted by AI.

Chapter IX — Review And Validity

Article 20  Review

This Policy shall be reviewed every year (or more frequently given the rapid pace of AI development) or at any time in the event of a significant shift in the AI landscape or applicable regulations.

Article 21  Validity

This Policy is effective from the date of enactment.

Established in: JAKARTA
Date: 1 January 2023.

(Signature)


Executive Director, A R K Foundation

(Signature)


Board Chairperson, A R K Foundation