Close Menu
    Sell your car stress-free. Peach Cars will handle all the paperwork and negotiations - Khusoko - East African Markets

    Khusoko AI Policy

    With the increasing adoption and usage of generative Artificial Intelligence (AI) tools and resources for content production, it is crucial to establish a set of guiding principles that ensure the responsible use of these technologies within Khusoko’s news value chain.

    This policy will be regularly updated to reflect the evolving nature of AI technologies, applicable law, and their applications, always adhering to Khusoko’s editorial guidelines and the Media Council of Kenya’s Code of Conduct for Media Practice, 2025, issued under the Media Council Act, 2013. ▸ (Updated reference: the Media Council’s Code of Conduct was revised in 2025 and now speaks directly to AI use in newsrooms — see below.)

    ▸ Legal position: This policy reflects Kenyan law as at the date of last review. Kenyan law and regulation on AI is still developing, and this policy will be reviewed periodically to stay current with legal and regulatory change, in line with the applicable laws set out throughout this document.

    What is Artificial Intelligence?

    While there’s no single, universally accepted definition of AI, the London School of Economics defines it as “a collection of ideas, technologies, and techniques that relate to a computer system’s capacity to perform tasks normally requiring human intelligence.”

    At Khusoko, we consider AI to encompass the technologies and processes, including algorithms and other tools, to enhance our workflow efficiency and deliver relevant content to our diverse audience. This includes using Large Language Models (LLMs), AI-powered image or video generation tools, automated transcription and translation tools, and AI-powered analytics and recommendation systems.

    While generative AI offers the potential to improve newsgathering and dissemination, it also presents potential risks to journalistic credibility, legal liability, and our unique relationship with our audience.

    Khusoko’s Commitment to Responsible AI Use

    We are committed to ensuring that our use of generative AI aligns with our core editorial values and applicable Kenyan law. Key commitments include:

    • Transparency: We believe in full transparency with our audience. We will always clearly disclose the use of AI tools in the creation or enhancement of our content, whether it’s in articles, images, audio, or videos.
    • Human Oversight: Human oversight is paramount. AI tools will be used as assistive technologies, with journalists and editors maintaining control over all stages of the content creation process. Every piece of AI-assisted content will undergo rigorous human review to ensure accuracy, fairness, and adherence to Khusoko’s editorial standards.
    • ▸ Accountability: A named senior editor is responsible for the implementation and outcomes of AI use across the newsroom (see “Accountability and Guidance” below), consistent with the four-step approach recommended by the UN and African Commission’s Joint Declaration on AI, Freedom of Expression, and Media Freedom (October 2025).
    • ▸ Proportionality and risk assessment: Before adopting a new AI tool or use case, we will assess the risk it poses — to accuracy, to source confidentiality, to data protection, and to our legal exposure — proportionate to how the tool will be used.

    Commitment to Quality

    While leveraging AI technologies, Khusoko remains steadfast in its commitment to:

    • Maintaining the highest standards of journalistic integrity.
    • Protecting source confidentiality and personal privacy.
    • Ensuring factual accuracy through rigorous human verification.
    • Preserving our unique voice and maintaining a strong relationship with our audience.
    • ▸ Avoiding bias and unfair representation. AI tools are frequently less accurate and more prone to bias in African, Kenyan and Swahili-language contexts than in English-language, Global North contexts, because they are trained on less representative data. Journalists and editors will apply extra scrutiny to AI output involving Kenyan people, places, languages (including Swahili and local languages), and current events, and will not treat AI-polished language as a proxy for accuracy.

    Editorial Integrity

    Editorial integrity is paramount. We will carefully evaluate all proposed uses of AI, prioritising journalistic purposes and acknowledging the potential risks associated with AI-generated content, such as plagiarism, hallucination (AI confidently producing false or fabricated information), or copyright infringement.

    Guidelines for AI Usage

    When considering the use of AI tools, team members should ask themselves:

    • How will AI be used specifically for this task?
    • Is this the most efficient and appropriate approach?
    • What is the clear journalistic purpose of this work?
    • How will the results be thoroughly fact-checked (where applicable)?
    • Is this for brainstorming, research, publishing, or another purpose?
    • Does the material involve sensitive topics or potentially harmful content?
    • Does the prompt contain any sensitive or private information? (e.g., revealing anonymous sources, names, dates of birth, addresses, names of minors)
    • ▸ Could the output mislead our audience if it isn’t clearly labelled, or if it depicts a real, identifiable person?
    • ▸ Does the underlying legal position on this use of AI remain unsettled (for example, copyright ownership of AI-generated output, or automated decision-making)? If so, treat the more risk-averse option as the default.

    Use-Case Guidance

    ▸ This section replaces the simple “Permitted/Prohibited” list with more specific guidance by use case, so that journalists know not just whether a use is allowed, but what safeguards apply. Where a use case is not listed here, default to the general Permitted/Prohibited rules below and consult an editor.

    Use case Guidance Key legal considerations
    Research & background gathering Permitted with human verification. Prompting AI for article ideas is discouraged. —
    Web scraping (e.g., monitoring government sites, company filings, regulatory announcements) Permitted for publicly accessible data only, limited to what is reasonably necessary for a story, and only where less invasive options (such as an official API) are not available. Do not bypass CAPTCHAs, paywalls, or other access controls. Data Protection Act, 2019; Computer Misuse and Cybercrimes Act, 2018/2024; Copyright Act, 2001
    Transcription of interviews and meetings Permitted, provided it is disclosed to all participants before the interview starts, verified for accuracy by a human editor before any quote is published, and never used for interviews involving confidential or off-the-record sources unless the tool is fully offline/internal. Transcripts are retained only as long as necessary and access is limited to relevant staff. Data Protection Act, 2019(source protection, data minimisation)
    Translation (e.g., Swahili/English or other local languages) Permitted as a first draft only. All AI-translated content must be reviewed by a fluent human speaker before publication and clearly labelled as AI-assisted where substantial. Extra scrutiny applies given known accuracy gaps in under-represented languages. Defamation Act, 1970
    Copy-editing / proofing Permitted for spelling, grammar and style only — not for structural or substantive edits, and never for changing words that carry legal significance (e.g., removing “alleged”). All AI-assisted copy is reviewed by a human editor before publication. Do not input copy containing information that could identify a confidential or vulnerable source. Defamation Act, 1970; Data Protection Act, 2019
    Fact-checking AI tools may assist but never replace human fact-checking, and must not be the sole basis for labelling a claim true or false, particularly for claims relating to Kenya or East Africa given known regional/linguistic bias in AI training data. All output undergoes human review. Defamation Act, 1970; Media Council Act, 2013
    Image/video generation Permitted only for clearly labelled illustrative content (e.g., non-news, opinion, or lifestyle pieces) — not for depicting real, identifiable people, real events, or anything that could reasonably be mistaken for an actual photograph or footage. See “Prohibited Use Cases” below. Copyright Act, 2001; Defamation Act, 1970; Consumer Protection Act, 2012
    Audience analytics / personalisation Permitted where based on aggregated or explicitly-consented data, limited to what is necessary, and where any automated recommendations are balanced with non-algorithmic options (e.g., “Most Read”, editor’s picks) to avoid filter bubbles. Personal data used for this purpose must not be retained longer than necessary. Data Protection Act, 2019(including the right not to be subject to a decision based solely on automated processing, s.35)
    Comment/content moderation (where applicable) AI may assist with initial flagging, but final decisions to remove or suppress user content are made or reviewed by a human, with a clear process for users to challenge a moderation decision. Constitution of Kenya, 2010 (freedom of expression, equality); National Cohesion and Integration Act, 2008; Data Protection Act, 2019

    Permitted Use Cases (general)

    AI may be used with human verification for:

    • Research and background information gathering (Note: prompting AI for article ideas is strongly discouraged)
    • Image illustrations (clearly labelled as AI-generated)
    • Audio/video editing and enhancement
    • Translation and transcription
    • Data analysis and visualisation
    • ▸ Web scraping of publicly available data, subject to the safeguards above

    Prohibited Use Cases

    The following uses of AI are strictly prohibited:

    • Using AI to generate or edit news stories.
    • Generating lifelike images or videos of real people.
    • AI-based fact-checking (as the sole or final check).
    • AI-generated or manipulated photos of real events.
    • Using AI tools to handle sensitive or confidential information, including information that could identify a confidential source, a minor, or a vulnerable individual.
    • ▸ Inputting classified, national-security-sensitive, or otherwise legally restricted information into any AI tool, particularly any tool hosted on an external/third-party server, where doing so could itself amount to an unlawful disclosure.
    • ▸ Using AI tools to access information behind a security control (e.g., CAPTCHAs, paywalls, login walls) without authorisation.

    Data Protection and Source Confidentiality

    ▸ New section — Khusoko’s original policy referenced protecting source confidentiality as a commitment but did not set out how. This section operationalises it.

    • Personal information (including that of sources, interviewees, and audience members) will not be input into third-party AI tools unless necessary for a clear editorial purpose, and, where practicable, will be anonymised first.
    • Copy, audio, or transcripts that identify or could identify a confidential or vulnerable source must not be input into any AI application, with particular care taken for investigative reporting.
    • Where a third-party AI tool is used, staff should be aware that data submitted to it may be transmitted to, processed, or retained by servers outside Kenya, and should consider this before submitting sensitive material — consistent with the safeguards on cross-border data transfer in the Data Protection Act, 2019.
    • Audience data collected via AI-powered analytics tools will only be processed for stated, legitimate purposes, retained no longer than necessary, and, wherever possible, aggregated and anonymised.
    • This section operates alongside, and does not replace, Khusoko’s Privacy Policy.

    Copyright

    ▸ New section, given ongoing legal uncertainty in Kenya about copyright ownership of AI-generated content.

    • Kenyan copyright law currently requires “sufficient human intervention” for a work to attract copyright protection (Copyright Act, 2001; Copyright Tribunal, Aryeh Movement Limited v. Cynthia Beldina Akoth Okello (2025)). Fully automated, unedited AI output may not be protected by copyright at all, meaning it could be freely reused by others.
    • Journalists and editors should therefore ensure any AI-assisted output that Khusoko intends to claim ownership over reflects substantive human authorship, editing, or creative input — not a lightly-edited AI draft.
    • Where possible, Khusoko will use AI tools whose terms of service vest ownership of generated output in Khusoko.
    • Content should not be generated by feeding another outlet’s published work into an AI tool to produce a “rewritten” or “summarised” version for publication — this raises both copyright and plagiarism concerns.

    Guidelines for Contributors

    • All contributors are expected to adhere to Khusoko’s AI policy.
    • All use of AI tools in the content creation process must be disclosed.
    • Editors reserve the right to:
      • Inquire about AI usage in the submitted content.
      • Request original, unedited drafts for review.
      • Make final decisions regarding the acceptability of AI-assisted content.
    • Contributors are strongly encouraged to maintain their unique writing style and leverage their expertise on the subject matter.

    Content Labelling

    Readers can expect clear labelling of AI-enhanced content:

    • AI-generated images/AI-assisted audio or video content will be clearly labelled.
    • When AI assistance is substantial in content creation, an editor’s note/label will be included.
    • If and when any social media post/article is created wholly by AI, editors will clearly state so in the title.
    • Any use of AI in the generation of content from advertisers/partners must be disclosed to the Sales/Marketing/Branded Content desk. Decisions about labelling for sponsored content will be made by our editorial team in accordance with this policy.
    • ▸ Where AI-generated or AI-assisted images or video are published, Khusoko will include an advisory that such content may contain inaccuracies, and will provide readers with a clear channel to flag suspected errors or rights concerns for prompt editorial review.

    Accountability and Guidance

    This policy provides a clear framework for the responsible and ethical use of AI technologies within the Khusoko newsroom. By adhering to these guidelines, we can ensure that AI is a valuable tool for enhancing our journalism while upholding the highest standards of accuracy, integrity, and transparency.

    ▸ In line with emerging international guidance for newsrooms (including the UN/OSCE/African Commission Joint Declaration on AI, Freedom of Expression and Media Freedom, October 2025), Khusoko will:

    • Designate a senior editor as accountable for the implementation and outcomes of AI use in the newsroom.
    • Conduct a basic risk assessment before adopting any new AI tool for editorial use.
    • Consider a prospective AI vendor’s data-handling and privacy practices before adoption.
    • Review this policy periodically against developments in Kenyan law, regulation, and industry best practice, and update it accordingly.

    Applicable Kenyan Laws

    ▸ New reference section, for internal use, mapping this policy to the Kenyan legal framework it is designed to help Khusoko comply with. This is not exhaustive and is not legal advice — newsroom staff with specific legal questions should escalate to an editor or seek legal counsel.

    This policy is informational and does not constitute legal advice. It draws on the Thomson Reuters Foundation/TrustLaw publication “Legal Guidelines on the Use of AI in Newsrooms (Kenya and Tanzania)” (July 2026), adapted for Khusoko’s own editorial practice.

    Updated 25 September 2026