A Powerful Tool in a Sensitive Beat
Artificial Intelligence (AI) is rapidly becoming part of the modern newsroom. For journalists covering Ghana’s Nuclear Power Programme, however, the technology presents a particularly delicate proposition: it can make reporting faster and more accessible, while an unchecked error can distort public understanding of radiation, nuclear safety or emergency preparedness. The challenge is therefore not whether journalists will use AI, but whether they will use it responsibly enough to preserve accuracy, accountability and public trust.

Speaking via a virtual workshop for media professionals by Nuclear Power Ghana on Thursday, September 17, 2026, General Manager, External Communications at the Electricity Company of Ghana (ECG), Dr. Charles Nii Ayiku Ayiku, describes responsible AI use in journalism as a way of strengthening “speed, accuracy, ethics and risk communication” in the nuclear landscape. His central warning is simple:
“AI can assist every stage of reporting, but it should own none of them.”

AI as Both Story and Tool
Artificial Intelligence encompasses technologies including machine learning, generative AI, large language models and multimodal systems. Dr. Ayiku Ayiku sums it up when he said that in nuclear reporting, AI is simultaneously “something journalists report about and something they report with”.
AI itself is already relevant to the nuclear landscape, with potential applications ranging from predictive maintenance and anomaly detection to safety, reliability, emergency preparedness, and radiation and environmental monitoring.

For journalists, the technology can also assist with research, analysis, interviews, writing, editing, publishing, and monitoring. But assistance is not authorship. A newsroom must retain human responsibility for every consequential claim.

Making Complex Nuclear Issues Understandable
Nuclear journalism presents a steep technical learning curve. Concepts involving radiation, reactor systems, radioactive waste, dose, safety standards and environmental monitoring can be difficult to understand without specialist knowledge. AI can provide a useful first orientation. It can explain technical concepts in plain language, identify unfamiliar terminology and help a reporter develop background questions. That is potentially the fastest legitimate newsroom benefit.
“Use AI to get oriented, never get quoted,” Dr. Ayiku Ayiku noted.

The same principle applies to investigative and data journalism. AI can help extract information from large document collections, organise evidence, identify inconsistencies and surface patterns in datasets. But a pattern is only a reason to investigate. It is not proof.
For Dr. Ayiku Ayiku:
“AI surfaces leads; journalism establishes truth.”

Preparing Better Questions
AI can also improve preparation before a technical interview. A reporter can use it to map an expert’s field, identify standard debates, and generate questions that move beyond the obvious. The objective is to enter the interview knowing enough to ask the second and third questions, not to replace the expert.
Transcription and translation offer another major time-saving opportunity. AI can rapidly turn interviews, meetings, and documents into searchable text and improve accessibility.

Yet this comes with a predictable weakness: transcription tools can struggle precisely where nuclear reporting becomes most technical. A misheard radiation unit, scientific term, facility name, or quotation can alter meaning.

When AI Gets the Science Wrong
The most serious risks arise when plausible-sounding AI output is mistaken for verified information. Generative AI can hallucinate facts, fabricate citations, produce outdated information, invent quotations, or generate incorrect figures. It may also confuse technical terms that sound similar but describe very different phenomena. In nuclear reporting, such errors are not merely cosmetic.
An incorrect radiation term, unit, measurement, or quotation can create unnecessary alarm or false reassurance. A technically defensible sentence can also give the public a misleading overall impression. That, too, is an accuracy failure.

Protecting Sources and Confidential Information
The risks extend beyond factual accuracy. Journalists handling unpublished documents, whistleblower material, embargoed information, or sensitive sources must understand where AI tools send and store data. Pasting confidential material into an external AI system without appropriate safeguards can compromise source protection. The same caution applies to unpublished investigations and sensitive security-related information. AI should never become a shortcut around established journalistic duties: source protection, attribution, fairness, right of reply, editorial judgement and public-interest decisions remain human responsibilities.

The Verification Chain
AI-generated material should therefore be treated as a starting point, not a finished fact. Before publishing a consequential nuclear claim, journalists should:

1. Return to the original source, including official documents, scientific papers, regulatory records or direct statements.

2. Verify every consequential claim, rather than accepting an AI summary.

3. Check numbers, units, dates, and names against authoritative sources.

4. Examine uncertainty and distinguish established facts from estimates, projections, or disputed claims.

5. Verify quotations against the original recording, transcript or document.

6. Check citations independently to ensure they actually exist and support the claim.

7. Seek expert clarification where technical meaning remains uncertain.

8. Fact-check the final story, including AI-assisted material that appears routine.

Trust Is the Asset
The nuclear newsroom must recognise that trust is the asset. Speed matters, and AI can deliver it. But where misinformation can influence public behaviour and perceptions of safety, speed without verification can become a liability. The responsible journalist uses AI to work faster, understand complex subjects and ask better questions, while keeping accuracy, accountability, source protection, professional judgement and public safety firmly under human control. The question is no longer whether journalists will use AI. It is whether they will use it responsibly enough to preserve the public trust on which credible nuclear reporting depends.

A Safer AI-Assisted Workflow
A practical newsroom workflow can begin by defining the story, gathering authoritative sources, using AI for orientation, and identifying unanswered questions and evidence gaps.
For example, if a reporter encounters a claim about “unusual radiation readings near a facility,” AI may help explain possible terminology and generate questions. It cannot establish that an unusual reading occurred, what caused it, whether it represents a health risk, or whether the claim is accurate.
Those questions require documents, measurements, qualified experts, regulators and independent verification.
Ultimately, the nuclear newsroom must recognise that trust is the asset. Speed matters. AI can deliver it. But in a beat where misinformation can influence public behaviour, political debate and perceptions of safety, speed without verification can become a liability.

Legal Consequences of Fake News and Defamation
As Artificial Intelligence reshapes newsrooms, journalists are being urged to ensure that speed and convenience do not come at the expense of accuracy.
Lecturer at the University of Media, Arts and Communication Institute of Journalism (UniMAC-IJ), Lawyer Zacharia Tanko Musah, has cautioned journalists to verify every piece of AI-generated information before publication. Speaking during a virtual workshop organised by Nuclear Power Ghana on September 18, 2026, he warned that unverified or defamatory content could expose both journalists and media organisations to legal consequences.
He advised:

“Ask AI for the sources of the information it provides and verify those sources before using the information, don’t use any data from AI if it cannot provide sources”.

Journalism Still Demands Accountability
This presupposes that managerial pressure or newsroom directives do not remove a journalist’s individual responsibility for published material, making the right of reply and the rule,

“No verification, no publication,” essential safeguards.
This implies that AI should remain a support tool, not an unquestioned authority. The final responsibility for what reaches the public rests with the journalist and the media organisation, making rigorous verification not optional, but essential to credible and responsible journalism.

The Path Forward
In this discourse, the responsible journalist therefore uses AI to work faster, think more broadly and ask better questions, while keeping truth, accountability and public safety firmly under human control.