ACCRA, Aug. 24, 2026 — MTN Ghana has trained journalists and representatives of Civil Society Organisations (CSOs) on the responsible use of Artificial Intelligence (AI) in newsroom and advocacy work as part of activities marking the telecom giant’s 30th anniversary in Ghana.
The training, held at MTN House in Accra and online, was themed “AI for Media & Civil Society” and facilitated by Samuel Adranyi, Chief Executive Officer of Afrilogic Solutions.
The training emphasised that AI tools should serve primarily as first-draft assistants and research accelerators, rather than sources of facts.
Participants were introduced to the basic workings of large language models, with facilitators explaining that AI converts words into tokens and numbers before predicting the next token in a sequence. They stressed that the fundamental objective of the technology is prediction, not truth.
Journalists were also taken through a four-step framework for safely integrating AI into their workflow: Delegate, Demand Evidence, Verify, and Decide.
Under the Delegate principle, participants were encouraged to assign AI only tasks that can be performed reliably without independent human intervention, such as organising disorganised source material, summarising notes and preparing first drafts.
Tasks requiring human verification, however, must remain under human control. These include identifying a suspect or victim before official confirmation, determining whether a claim is defamatory, assessing source credibility, authenticating images or videos, and making the final decision to publish, hold or spike a story.
A key concept introduced during the training was the Provenance Ledger, which requires every sentence in an AI-assisted draft to be classified as SUPPORTED, INFERRED or NEW before publication.
“SUPPORTED” refers to information directly stated in the source; “INFERRED” refers to information reasoned from the source; while “NEW” refers to information introduced by AI that is absent from the original source.
Facilitators stressed that every “NEW” item must either be deleted or independently sourced before publication, regardless of deadlines.
The central message of the training was that AI can draft the sentence, but a human must take responsibility for it. Where there is uncertainty, participants were urged to exercise caution and ensure that AI-generated content never overrides editorial judgement.
Participants were further introduced to a “Routine Rule” for determining what information may safely be entered into AI platforms.
According to the guidance, public information may be processed using mainstream AI tools, while sensitive information should only be handled through offline or privacy-committed tools. Confidential information, including sources, leaked documents and beneficiary data, must not be entered into consumer-facing AI tools.
The training forms part of efforts to strengthen AI literacy among Ghanaian journalists and civil society actors as newsrooms and advocacy organisations increasingly integrate generative AI into their daily operations.
Source: Robertha Bubune Fiatorwogbor
