AI Policy
Artificial Intelligence Policy
Centre for Investigative Journalism Malawi (CIJM)
Version: 1.0
Adopted: April 2026
Review Date: April 2027
Policy Owner: Executive Director,
CIJM
Applicability: All CIJM staff,
editors, journalists, freelancers, contributors, and institutional partners
The Centre for Investigative Journalism Malawi (CIJM) exists to hold power to account through independent, accurate, and impactful investigative reporting in the public interest. Since its founding, CIJM has upheld the highest standards of journalistic integrity. As artificial intelligence (AI) technologies become an increasingly embedded feature of modern newsrooms, CIJM recognises both the opportunities and the risks they present.
Globally, 81% of journalists are already using AI in their work, yet only 13% operate under established AI policies. In the Global South, the figure is even starker; as of late 2024, approximately 80% of journalists surveyed by the Thomson Reuters Foundation reported that their newsrooms have no AI policy. CIJM rejects this vacuum. This policy is designed to ensure that any use of AI at CIJM serves, rather than supplants, the rigorous journalism for which the organisation is known.
This policy is grounded in the following international frameworks and standards:
- The Paris Charter on AI and Journalism (RSF, November 2023)
- UNESCO’s Recommendation on the Ethics of Artificial Intelligence (2021) and its AI and the Future of Journalism issue brief (2024)
- The Council of Europe Guidelines on Responsible AI in Journalism (CDMSI, November 2023)
- The Alliance for Audited Media (AAM) Eight Pillars of Ethical AI
- The Malawi Data Protection Act, No. 3 of 2024
- Best practices from the International Consortium of Investigative Journalists (ICIJ) on AI-assisted investigations
This is a living document. It will be reviewed annually or whenever significant developments in AI technology or applicable Malawian law necessitate an update.
1.1 Purpose
This policy establishes clear, enforceable rules governing the use of artificial intelligence tools and technologies in all aspects of CIJM’s work. Its purpose is to:
- Protect CIJM’s editorial independence, credibility, and public trust
- Ensure AI is used as a supplement to, not a replacement for, human journalistic judgment
- Safeguard sources, whistleblowers, and sensitive investigation data from AI-related security risks
- Align CIJM’s practices with Malawian law and international ethical standards for journalism
- Provide practical guidance so that CIJM staff can confidently harness AI’s benefits while mitigating its risks
This policy applies to:
- All full-time and part-time CIJM employees, including editors, journalists, researchers, data analysts, and administrative staff
- All freelancers, stringers, and independent contributors working on CIJM assignments
- Institutional partners, training facilitators, and consultants engaged by CIJM
- Any person or entity accessing CIJM systems, data, or editorial workflows
This policy covers all AI tools and systems, including but not limited to: generative AI text tools (e.g., ChatGPT, Gemini, Copilot), image generation tools (e.g., DALL-E, Midjourney), transcription tools, translation tools, AI-assisted data analysis platforms, automated fact-checking systems, and machine learning tools used in document analysis.
Section 2: Governing Principles
CIJM’s use of AI shall at all times be governed by the following core principles, which are aligned with the Paris Charter on AI and Journalism and the UNESCO Ethics of AI Recommendation:
2.1 Human Agency and Editorial Primacy
AI systems are tools that assist, inform, and augment human journalism. They do not replace the editorial judgment, ethical reasoning, investigative instinct, or moral responsibility of CIJM’s journalists and editors. All final editorial decisions, on story selection, framing, publication, and retraction, remain exclusively with human editors. As the Paris Charter states, human decision-making must remain central to all aspects of journalism.
CIJM’s reputation rests on the accuracy of its reporting. AI systems are not capable of independently evaluating, critically combining, or verifying information. Every piece of information generated, synthesised, or surfaced by an AI tool must be independently verified by a human journalist before it is relied upon in any story. The burden of verification cannot be delegated to any AI system.
CIJM will be open with its audiences, sources, and the public about when and how AI has been used in its journalism. Transparency is not merely an internal standard. it is an external commitment that forms part of the organisation’s accountability to the public it serves.
Clear lines of accountability exist for all AI-assisted work at CIJM. Responsibility for accuracy, fairness, and originality always rests with the human journalist and editor who produced, edited, or authorised the content, not with the AI tool or its developer.
2.5 Fairness and Non-Discrimination
AI systems have been documented to perpetuate bias based on gender, race, ethnicity, age, and ability. CIJM will actively guard against such biases in any AI output it uses. The organisation’s longstanding commitment to fair, equitable, and representative reporting extends fully to AI-assisted journalism.
2.6 Source Protection and Confidentiality
The protection of sources and whistleblowers is a non-negotiable principle of investigative journalism. AI tools must never be used in ways that risk exposing source identities, compromising sensitive investigation data, or undermining the confidentiality of communications. This principle takes precedence over convenience or efficiency.
CIJM’s editorial independence shall not be compromised by reliance on proprietary AI systems whose operators may have conflicting commercial, political, or other interests. The organisation will critically assess all AI tools for potential bias embedded in their training data before adoption.
2.8 Privacy and Data Protection
All use of AI at CIJM must comply with the Malawi Data Protection Act, No. 3 of 2024, which establishes the Malawi Communications Regulatory Authority (MACRA) as the official Data Protection Authority. CIJM will not input personal data of sources, subjects of investigation, or members of the public into AI tools without a lawful basis for doing so under the Act.
Section 3: Permitted and Prohibited Uses of AI
The following uses of AI are permitted at CIJM, subject always to the principles in Section 2 and the safeguards in subsequent sections:
Research and Investigation Support
- Using AI tools to analyse and identify patterns in large datasets, leaked documents, financial records, procurement data, and public registers, consistent with ICIJ’s established practice in major investigations.
- Using machine learning to classify, tag, and search document sets in investigations involving high-volume leaks
- Using AI-powered Natural Language Processing (NLP) tools for sentiment analysis, entity extraction, or summarisation of large volumes of text as a starting point for human analysis
- Using AI tools to generate research leads, identify story angles, and map networks of interest, always subject to human verification
Workflow and Production Support
- Transcription of interviews and public proceedings, with human review and correction before use
- Translation of documents into English or local languages, with human verification by a competent speaker
- Generating headline variations and social media post suggestions, subject to editorial approval
- Automated captioning of multimedia content, with human correction
Audience Engagement and Administration
- AI-assisted SEO optimisation of published content
- Content tagging and categorisation for the CIJM website
- Administrative tasks such as scheduling, calendar management, and email drafting, clearly labelled as AI-assisted in internal systems
The following uses are expressly prohibited at CIJM:
- Using generative AI to write, ghost-write, or substantially draft any published news story, investigation, analysis, or editorial, in whole or in part, without clear disclosure and approval by the Executive Director
- Publishing AI-generated photographs, images, or illustrations as authentic documentary evidence of events
- Inputting names, contact details, locations, or identifying information of sources, confidential tipsters, or whistleblowers into any AI tool
- Uploading unpublished investigation documents, sensitive source material, or confidential interview notes into AI systems, especially cloud-based tools where data may be used for model training
- Using AI-generated content to fabricate quotes, reconstruct scenes, or simulate events not independently verified
- Using AI-deepfake or synthetic media tools to create or alter audio-visual content for publication
- Relying on AI fact-checking outputs as the sole verification mechanism for any claim in a published story
Section 4: Disclosure and Labelling Standards
Transparency with CIJM’s audiences is a hallmark of its editorial integrity. The following disclosure standards are mandatory:
- Any published story in which AI tools were used for data analysis, document processing, or research assistance shall carry a note in the methodology section or article footer explaining how AI was used and what human verification was applied.
- Any AI-generated illustration or image used on the CIJM website or in publications shall be clearly labelled: “Illustration generated using artificial intelligence.”
- Transcriptions produced with AI assistance shall note the tool used and that the transcript was reviewed and corrected by a human journalist.
- AI-assisted translations used as primary source material shall disclose the translation method and note that it was verified by a human speaker.
4.2 Internal Disclosure Obligations
- Journalists must declare AI tool use to their assigning editor at the point of submission.
- Editors must document AI use in the story’s internal editorial record.
- Freelancers and contributors must disclose AI tool use to their CIJM commissioning editor before submission, per the Contributor Guidelines in Section 9.
4.3 Audience Engagement on AI Use
CIJM will publish a public-facing AI Use Statement on its website (www.investigative-malawi.org/ai-policy), updating it whenever this policy is revised, to maintain open communication with its audience about its AI practices.
Section 5: Source Protection and Investigative Security
Given CIJM’s mandate as an investigative journalism organisation operating in a context where sources face real risks of retaliation, this section carries heightened importance.
5.1 Data Security Requirements for AI Tools
- AI tools used in live investigations must be evaluated for data security before adoption. Cloud-based tools that retain user inputs for model training are presumptively prohibited for use with investigation data.
- Sensitive investigation data shall only be processed using AI tools that have been assessed and approved by CIJM’s AI Oversight Committee (see Section 7), and where data processing agreements are in place consistent with the Malawi Data Protection Act.
- Where AI tools are used to process leaked or confidential documents, journalists must use secure, air gapped, or locally-run systems wherever feasible.
- No AI tool shall be used to attempt to identify, de-anonymise, or profile anonymous sources or tipsters.
- AI-generated analysis of communications, social media, or metadata that could lead to the identification of a source is prohibited.
- Journalists who receive tip-offs through digital channels shall not input communication details into AI systems for any purpose.
- Staff are prohibited from connecting to AI tools via unsecured public networks when handling investigation data.
- CIJM will incorporate AI security risks into its existing digital security training programme.
Section 6: Accuracy, Verification, and Editorial Standards
6.1 The Verification Imperative
Generative AI systems are not capable of independently verifying information and are known to produce plausible-sounding but factually incorrect outputs, a phenomenon widely documented as “hallucination”. Every factual claim surfaced, suggested, or supported by an AI tool must be independently verified through primary sources before publication.
6.2 Fact-Checking Protocol for AI-Assisted Content
When AI tools have contributed to any part of the reporting or research process, the responsible journalist must:
- Cross-reference all facts, statistics, names, dates, and attributions against primary or authoritative sources
- Verify that no AI-generated quote has been attributed to a real person who did not say it
- Confirm that no AI-generated data analysis conflicts with the underlying raw dataset
- Submit AI-assisted work to the standard editorial review process with a declaration of AI use
6.3 Corrections and Accountability
Where an error is found to have resulted from unchecked AI output, CIJM will:
- Publish a timely correction per its standard corrections policy
- Investigate how the error passed the verification process
- Update staff training or this policy as required to prevent recurrence
- The responsible journalist and editor bear personal accountability for the error, not the AI tool
Section 7: Governance and AI Oversight Committee
7.1 Establishment of the AI Oversight Committee
CIJM shall establish an AI Oversight Committee (AOC) to govern the implementation of this policy. The AOC shall comprise:
- The Executive Director (Chair)
- The Investigations Editor
- A designated AI Policy Lead (a senior journalist with AI literacy training)
- One representative from CIJM’s technology or data team
- An external ethics advisor or civil society representative (to be appointed by the Board)
7.2 Mandates of the AI Oversight Committee
The AOC is responsible for:
- Approving or rejecting AI tools for use in CIJM’s editorial and operational workflows
- Monitoring compliance with this policy across all teams and projects
- Receiving and investigating complaints about AI misuse
- Reviewing this policy annually and recommending amendments
- Keeping CIJM staff informed about developments in AI technology and regulation
- Liaising with peer organisations, including ICIJ, Thomson Reuters Foundation, and African journalism bodies, on emerging AI standards.
The AOC shall appoint a Designated AI Policy Lead whose responsibilities include:
- Serving as the first point of contact for staff with questions about AI tool use
- Maintaining a register of approved AI tools and their terms of use
- Coordinating training sessions on AI use and ethics
- Monitoring AI tool developments and flagging changes in tool providers’ data policies that may affect CIJM’s source protection obligations
Section 8: Training, Capacity Building, and Culture
8.1 Mandatory AI Literacy Training
All CIJM staff, including editors, journalists, researchers, and administrative personnel, shall undergo mandatory AI literacy training as part of their onboarding and on an annual basis thereafter. This training shall cover:
- How generative AI systems work, including their limitations and known failure modes
- CIJM’s specific rules for AI use, including what is permitted and prohibited under this policy
- How to apply AI tools responsibly in investigative reporting and data analysis
- Source protection in the AI era: secure use of AI tools, data handling risks, and digital security
- How to identify and mitigate AI-generated bias and misinformation
- Disclosure and labelling obligations
8.2 Collaboration with Experts
CIJM will build relationships with data scientists, AI ethics researchers, digital rights organisations, and technologists to ensure its AI practices remain grounded in current knowledge. This includes engagement with:
- The International Consortium of Investigative Journalists (ICIJ) on AI-assisted investigative tools.
- The Thomson Reuters Foundation’s AI-ready newsroom guidance.
- Regional bodies including the African Journalism Centre and Wits Centre for Journalism.
8.3 Experimentation and Innovation
CIJM recognises that the responsible use of AI can significantly enhance the reach and impact of its investigations. Staff are encouraged to identify new AI applications that serve the organisation’s mission and to bring these to the AOC for assessment. Innovation shall not be treated as automatically suspect; the guiding test is whether a proposed AI application serves CIJM’s public interest mandate while respecting the principles in Section 2.
Section 9: Contributor and Freelancer Guidelines
Freelancers, stringers, and contributors who produce work for CIJM are bound by this policy in full for any AI use related to their CIJM assignments.
9.1 Obligations of Contributors
- Contributors must disclose to their commissioning editor, in writing, any use of AI tools during the research, writing, or production of their assignment before submission.
- Contributors remain fully accountable for the accuracy and originality of the content they submit. AI use does not transfer liability to the AI tool or to CIJM.
- Contributors shall not input CIJM source names, contact information, or sensitive investigation material into any AI tool.
- Contributors who use AI for transcription, translation, or research assistance must ensure all outputs are human-verified and clearly labelled in their submission notes.
A contributor who violates this policy, including by submitting AI-generated content without disclosure, compromising source protection, or inputting confidential CIJM material into AI systems, may have their assignment terminated without payment and may be barred from future engagement with CIJM. Serious breaches will be referred to CIJM’s editorial leadership for further action.
Section 10: Intellectual Property and Copyright
10.1 Copyright in AI-Assisted Work
CIJM acknowledges the significant unresolved questions around copyright in AI-generated content, including the legal and ethical questions about AI systems trained on journalistic content without authorisation or compensation.
CIJM’s position is:
- Content substantially generated by AI systems does not automatically attract the same copyright protections as human-authored journalism.
- CIJM journalists and editors retain copyright in work they have substantially created, regardless of AI tool assistance.
- CIJM will not sell, license, or submit its published investigative content to AI training platforms without a board decision, proper legal review, and, where required, compensation agreements that protect the organisation’s interests.
10.2 Respecting Third-Party Intellectual Property
- Staff must not input third-party copyrighted text, images, or data into AI tools in ways that reproduce or republish that content without authorisation.
- AI-generated text must be reviewed to ensure it does not contain unattributed reproductions of copyrighted material.
- Any licensing or IP questions related to AI use must be escalated to the Executive Director.
Section 11: Complaints, Breaches, and Enforcement
Any staff member, contributor, or member of the public who believes this policy has been violated may report the concern to the Designated AI Policy Lead or directly to the Executive Director. Reports may be made confidentially.
11.2 Investigation of Breaches
Reported breaches will be investigated by the AI Oversight Committee. The investigation will:
- Assess whether a policy violation occurred
- Determine the impact on editorial quality, source safety, or reputational risk
- Recommend remedial action, ranging from additional training to disciplinary measures
11.3 Disciplinary Consequences
Deliberate or reckless violations of this policy, particularly those that compromise source protection, publish unverified AI content, or involve misleading the audience about AI use, will be treated as serious breaches of CIJM’s editorial code and may result in disciplinary action up to and including termination of employment or contract.
Section 12: Review and Iteration
This policy shall be reviewed:
- Annually, no later than 12 months from the date of adoption
- Ad hoc, whenever: (a) a major new AI capability or tool becomes relevant to CIJM’s operations; (b) applicable Malawian law changes; (c) a significant AI-related incident occurs at CIJM or a peer organisation that has lessons for CIJM; or (d) the AOC determines a review is warranted
All staff shall be notified of updates to this policy within five working days of adoption. The updated policy shall be published on CIJM’s website.
Appendix A: Approved AI Tools Register
The Designated AI Policy Lead shall maintain a living register of AI tools approved for use at CIJM, including the tool name, approved use cases, applicable limitations, and data handling assessment status. This register is available to all CIJM staff on request.
| Term | Definition |
| Artificial Intelligence (AI) | Computer systems designed to perform tasks that typically require human intelligence, including understanding language, recognising patterns, generating text or images, and making decisions |
| Generative AI | AI systems that generate new content — text, images, audio, video — in response to user prompts, including tools like ChatGPT, Gemini, Claude, and Midjourney |
| Large Language Model (LLM) | A type of AI trained on vast text datasets to understand and generate human language |
| Hallucination | The tendency of generative AI to produce confident-sounding but factually incorrect or fabricated information |
| Machine Learning | A branch of AI in which systems learn from data to improve performance on specific tasks over time |
| Natural Language Processing (NLP) | AI techniques for analysing, understanding, and generating human language |
| Source Protection | The journalistic obligation to protect the identity and safety of individuals who provide information to journalists in confidence |
| Data Controller | Under the Malawi Data Protection Act 2024, an entity that determines the purposes and means of processing personal data |
| AI Oversight Committee (AOC) | CIJM’s internal governance body responsible for overseeing compliance with this policy |
Appendix C: Key International References
| Framework | Year | Issuing Body | Key Contribution |
| Recommendation on the Ethics of AI | 2021 | UNESCO | Global standard on human-centred AI ethics; promotes transparency and freedom of expression |
| Paris Charter on AI and Journalism | 2023 | RSF + 16 partners | Ten principles protecting journalistic integrity in the AI era |
| CoE Guidelines on Responsible AI in Journalism | 2023 | Council of Europe | Guidance on aligning AI with freedom of expression and editorial mission |
| AI and the Future of Journalism | 2024 | UNESCO | Issue brief for newsrooms on governance, copyright, and editorial standards |
| Three Steps to an AI-Ready Newsroom | 2025 | Thomson Reuters Foundation | Practical guide for newsrooms identifying and mitigating AI risks |
| Malawi Data Protection Act | 2024 | Government of Malawi | Legal framework governing personal data processing, enforced by MACRA |
| AAM Ethical AI Framework | 2026 | Alliance for Audited Media | Eight pillars for responsible AI use in media organisations |
This policy was developed by the Centre for Investigative Journalism Malawi (CIJM). For questions or submissions, contact the AI Policy Lead at: cijm@investigative-malawi.org | www.investigative-malawi.org