By Pita Ligaiula
Pacific newsrooms risk becoming consumers of artificial intelligence systems shaped outside the region unless they build their own tools, rules and knowledge base, says Pasifika AI creator and Kanaky Tech founder Kevyn Wahuzue.
Wahuzue told PACNEWS that current AI systems often fail to distinguish between Pacific peoples and languages, while limited connectivity, training and newsroom resources could create another form of information inequality.
“To be honest, these models don’t ‘understand’ anything. They reproduce what they were trained on, and very little of that was written by Pacific people, in Pacific languages,” Wahuzue said.
He said one of the most common failures he had identified was the blending of Pacific cultures and languages.
“In my tests, when I asked for an answer in Cook Islands Māori, some models wrote New Zealand Māori instead.
“Related languages, different peoples. A model trained mostly outside the region treats the Pacific as one culture with one voice.”,” he told PACNEWS.
Wahuzue said Pacific institutions, academies and archives should be used to ground AI systems, rather than relying heavily on information from the open web.
He said Pacific communities should also have the power to correct systems and request the removal of information.
“Ground the answers in sources from the region: our institutions, academies, archives, not the open web.
“Keep each people distinct and named. And let the people who hold the knowledge correct the system, with the power to have something taken down,” he said.
Wahuzue pointed to Te Hiku Media in Aotearoa New Zealand as an example of Indigenous-led technology development, saying it had developed its own te reo Māori speech technology while retaining control of its data.
Wahuzue said AI could create another information divide in the Pacific, but argued that the cost of AI tools was not necessarily the main barrier.
“The tools are getting cheap. Pasifika AI runs on a daily budget of a few dollars,” he said.
He identified connectivity, training, limited newsroom time and, particularly, language as bigger challenges.
“In my tests, most models could not write Samoan, Tongan or Tahitian correctly, and the weaker ones don’t tell you.
“A newsroom publishing in English gets a strong assistant. A newsroom publishing in a Pacific language gets a weak one, or a dangerous one,” he said.
Wahuzue said Pacific news organisations should work collectively to develop AI skills, rules and tools.
“If Pacific newsrooms only consume AI, the decisions about what these systems know about us get made somewhere else.
“The region should pool it: shared training, shared rules, shared tools across newsrooms, the same way PACNEWS pools the news,” he said.
With AI making it easier to create convincing fake images, audio and video, Wahuzue said Pacific journalists needed stronger verification practices.
He identified three basic habits: verify the source, explain the verification process and never publish AI-generated depictions of real people or events as if they were authentic.
“First, verify the source, not just the file. Who sent it, where it came from, can you reach the person who recorded it,” he said.
“Detection tools help, but they lose the race: the fakes get better every month,” he told PACNEWS.
He said journalists should also tell audiences how they established whether material was genuine.
“Tell your audience ‘we reached the person who filmed this’ or ‘we could not confirm this.’ People trust a method they can see,” he said.
Wahuzue also warned against using AI-generated images of real people or events, even when presented as illustrations.
“Never use AI to create an image of a real person or a real event and publish it as if it were real, not even as an illustration.
“Once a generated face sits next to a real name, it travels without the caption,” he said.
He said Pacific journalists could also draw on the region’s close-knit communities when checking suspicious material.
“Pacific audiences have one advantage: small, connected communities. Someone always knows the person in the video.
“Use that network before you publish, not after,” Wahuzue said.
Wahuzue said responsibility for an AI-assisted story ultimately remained with the newsroom.
“The newsroom. Always,” he said when asked who should be accountable when AI-assisted reporting contains false or misleading information.
“The journalist is responsible for what they file, the editor for what they approve, the organisation for what it publishes. That chain doesn’t change because a machine was involved,” Wahuzue said.
He said blaming AI for an inaccurate story should not be accepted by news organisations.
“‘The AI got it wrong’ is the new ‘the intern did it.’ Readers won’t accept it, and they shouldn’t,” he said.
Wahuzue said technology providers should be accountable for claims they make about their products, but that responsibility for published journalism remained with the media organisation.
“The byline and the masthead carry the story. You can outsource a task. You can’t outsource the responsibility,” he said.
He said the same principle applied to Pasifika AI.
“For my own tool, that means my contact is on every guide and corrections come to me.
“If Pasifika AI gets something wrong, it’s my name on it, not the model’s,” Wahuzue said.
