Kenya - Media Reporting Guidelines

Overview

The draft Media Handbook for Reporting on Artificial Intelligence in Kenya was produced by a technical committee convened by the Media Council of Kenya (MCK) to provide practical guidance for journalists, editors and media organisations. It forms part of a suite of documents that include a Data Governance Guide for Media Practice, a Media Guide on the Use of Artificial Intelligence in Kenya and a Guide on the Use of Social Media and the Internet for Media Practice in Kenya. The handbook addresses dual concerns: how journalists should responsibly report about AI (companies, government use, societal impacts) and how newsrooms should ethically deploy AI tools in newsgathering, production and distribution. The draft responds to rapid adoption of AI tools in news ecosystems, the presence of disinformation enabled by synthetic media, and the need to preserve journalistic standards and public trust while enabling beneficial innovation. MCK publicly announced receipt of taskforce outputs in January 2024 and invited comments on the draft in January 2025 as part of a constitutionally-mandated public participation process; background on these steps is available via the Council's newsroom releases and press coverage. The handbook is explicitly practice-oriented, cross-referencing Kenyan legal obligations (notably data protection) and international best practice resources such as the UNESCO handbook on reporting AI.

Definitions

The draft provides concise definitions tailored for media practice. Key definitions include: "Artificial Intelligence (AI)" — computational systems that perform tasks commonly associated with human intelligence, including machine learning models, large language models (LLMs), generative models and decision-support systems; "AI-assisted content" — material where AI tools contributed to ideation, drafting, editing, audio/video synthesis, or visual generation; "AI-generated content" — outputs produced primarily by algorithmic systems without human editorial crafting; "Model provider/vendor" — entities offering AI models, services or APIs to media organisations; "Personal data" — defined in alignment with Kenya's Data Protection Act and denotes any information relating to an identifiable person; and "Explainability/interpretability" — the extent to which a model's operation and outputs can be meaningfully described to editors, subjects and the public. These operational definitions are designed to be usable in newsroom policy documents, editorial codes and training curricula.

Governance and Institutional Framework

The handbook situates MCK as the lead sector regulator and standard-setter for media ethics and practice in Kenya, recommending coordination with statutory bodies responsible for data protection, communications and consumer protection. It encourages media houses to adopt internal governance mechanisms — editorial AI policies, dedicated editorial AI officers, procurement checks, and cross-functional oversight committees comprising editorial, legal and technical staff. The handbook recommends mechanisms for public engagement, complaints handling and transparency reporting so that audiences can hold outlets accountable for AI use. The draft also outlines MCK's process for stakeholder consultation (open call for comments, technical workshops) and envisages an ongoing role for MCK in accreditation of AI training providers for journalists and in publishing model guidance on vendor audits. For regional and sectoral coherence the handbook recommends memoranda of understanding with the Media Council of Kenya's partner agencies, and periodic multi-stakeholder fora to align media practice with developments in data protection, broadcasting regulation and consumer safety.

Key Focus Areas

The handbook concentrates on several interlocking domains: (1) Transparency and Disclosure — requiring clear labelling of AI-generated or AI-assisted content and disclosure of significant editorial reliance on models; (2) Accuracy, Verification and Fact-Checking — prescribing rigorous verification standards for claims grounded in algorithmic outputs and a duty to disclose known limitations or uncertainties; (3) Data Protection and Consent — instructing journalists on lawful handling of personal data when training or querying models and on minimising unnecessary data exposure; (4) Editorial Oversight and Accountability — mandating human editorial control over final publication decisions and robust audit trails documenting AI use; (5) Explainability and Public Interest Assessment — recommending plain-language explanations of model roles in stories affecting public interest and guidance for how to evaluate harm and public benefit; (6) Security and Model Robustness — promoting cybersecurity hygiene, vendor security checks and controls against poisoning or manipulation; and (7) Skills, Training and Capacity Building — requiring continuous upskilling, specialist training and a pool of accredited trainers. The draft also highlights sector-specific sensitivities (e.g., health and finance reporting) where erroneous or biased AI outputs can cause acute harms and where extra safeguards — such as expert review and conservative editorial thresholds — are recommended. Across these areas the handbook references international toolkits and local legal obligations, urging balancing innovation with rights protection. The handbook stresses that labelling alone is insufficient: editorial processes, verification, redress mechanisms and structural safeguards are required to protect public trust.

Implementation Framework

MCK's draft proposes a multi-layered implementation approach. At the organisational level, media outlets are advised to adopt an "AI editorial policy" that documents permitted uses, disclosure templates, approval workflows and remediation steps for errors. Procurement procedures should include vendor due diligence for model provenance, training data practices and contract clauses addressing explainability, liability and data handling. At the editorial level, the handbook prescribes mandatory checks for AI-assisted investigative work, a requirement for human sign-off on AI-generated material, and special protocols for sensitive categories (elections, health, minors). MCK proposes to assist implementation by publishing template policies, organising accredited training sessions and maintaining a registry of accredited trainers and external technical auditors. The handbook recommends that larger outlets establish internal audit logs and versioning systems for AI prompts and outputs, and that outlets of all sizes maintain documented correction policies that explicitly cover AI errors. For capacity building, the draft foresees MCK-supported workshops, cross-sector secondments and an online resource hub, drawing on international reference materials such as the UNESCO reporting AI handbook to accelerate adoption.

Monitoring and Evaluation

The draft sets out monitoring and evaluation (M&E) mechanisms to track adoption and impact. It recommends periodic self-assessment checklists for media houses, a simple reporting template for MCK to collect anonymised compliance metrics (e.g., number of stories labelled AI-generated, number of corrections issued), and targeted audits for outlets using AI in high-impact reporting. MCK envisions collaborating with academic partners and independent auditors to spot systemic issues such as algorithmic bias or recurring verification failures. Outcome indicators proposed include improved disclosure rates, reduced incidence of unlabelled synthetic media, timeliness of corrections and increased journalist training participation. The handbook also recommends a complaints and appeals mechanism — integrated into existing MCK processes — with public reporting on unresolved complaints and remedial steps taken. Regular public and technical reviews (annual or biennial) are suggested to update the handbook in line with evolving technology and jurisprudence.

Penalties, Liability, and Appeals

As a draft guidance instrument anchored in MCK's regulatory remit over media standards, the handbook outlines corrective and disciplinary measures rather than criminal sanctions. Proposed sanctions include requirement to publish corrections or clarifications, suspension of accreditation for repeated or systemic breaches, mediation or adjudication under MCK's complaints process, and referral to other regulators (e.g., the Office of the Data Protection Commissioner or communications regulators) where statutory violations occur. The handbook encourages contractual allocations of liability in vendor agreements and recommends that outlets maintain insurance or indemnity arrangements for harms arising from AI deployment. An appeals process is described within MCK's established complaints and adjudication framework, providing transparency as to how editorial disputes involving AI will be handled. The draft emphasises proportionality and the primacy of remedial and corrective measures to preserve public interest and freedom of expression.

Relationship to Other Instruments

The draft explicitly cross-references Kenya's Data Protection Act and existing MCK Codes of Conduct for media practice, positioning the handbook as sector-specific guidance that supplements statutory obligations. It invites coordination with the Data Protection Commissioner on questions of lawful data processing for model training or investigative work. It also situates itself alongside broadcasting regulations and communications laws where applicable, recommending memoranda of understanding to avoid regulatory overlap. Internationally, the handbook draws on UNESCO and other global resources for reporting on AI and proposes to harmonise with cross-border standards on transparency, particularly for outlets reporting on transnational AI systems. The draft also recommends that media houses align contractual arrangements with vendor terms to preserve remedies and audit rights, and that MCK continue to liaise with judicial and legislative stakeholders to ensure coherence between soft guidance and hard law enforcement where necessary.

International Alignment

While tailored to Kenya's legal and media landscape, the handbook commits to international alignment by referencing global best practice instruments. It draws on the UNESCO handbook on reporting AI for pedagogical content, points to international journalism standards (press codes and fact-checking methodologies), and recognises international data protection norms when advising on cross-border data flows. The draft also incorporates principles from international AI ethics frameworks (transparency, accountability, safety) to ensure Kenyan practice can interoperate with regional and global media ecosystems. MCK signals intent to participate in international fora to keep the handbook aligned with technological change and transnational regulatory developments and to ensure Kenyan media obligations do not create competitive disadvantages nor compromise journalistic freedom.

Implementation Timeline

Sources and References