Justice Minister Paul Goldsmith has asked officials for advice on how to protect New Zealand’s electoral systems from the growing threat posed by artificial intelligence, with the general election now just four months away.
Goldsmith confirmed he expects that advice to land on his desk soon. The request follows a blunt warning from the country’s cyber security agency that the risks tied to the most advanced AI models are no longer theoretical. New Zealand goes to the polls on 7 November, and the question of whether cheap, convincing, machine-generated content could be used to mislead voters or disrupt the vote is now firmly on the table.
The trigger was fresh guidance from the National Cyber Security Centre, which sits inside the Government Communications Security Bureau. The centre joined its Five Eyes intelligence partners in Australia, Canada, the United Kingdom and the United States to issue a joint call to action on the dangers of frontier AI. In their assessment the agencies said the overall threat is real and immediate, language that is unusually direct for security officials who normally choose their words carefully.
Goldsmith framed his request as routine stewardship rather than alarm. Speaking to RNZ, he said the focus is on ensuring existing arrangements continue to be effective and responsive to the evolving threat environment. “I would always expect agencies to be actively considering their stewardship responsibilities to identify changes in the environment,” he said. He described protecting the integrity and security of New Zealand’s elections as a longstanding priority.
There is a limit to what the minister can do directly. The Electoral Commission is an independent body, and Goldsmith said he could not direct the commission on how it operated. That independence is a deliberate feature of the system. It keeps the party in government at arm’s length from the machinery that runs the vote, which matters a great deal when trust in the result is the whole point. It also means the response to AI cannot simply be ordered from the top down. It has to be worked out between the minister, the commission and the security agencies, each staying inside their own lane.
The commission, for its part, said its security practices already took frontier AI into account. It added that it was considering the implications of this newly identified threat on the readiness of its systems. In plain terms the commission is saying it has not been caught flat-footed, but it is also not claiming the problem is solved.
It helps to separate two very different problems that often get lumped together under the same headline. The first is protecting the physical and digital machinery of the election. That means the enrolment database, the systems that count and report votes, and the commission’s own public websites. This is a familiar cyber security job with familiar defences, and it is the part officials sound most confident about. Keeping hackers out of the count is hard work, but it is well understood work.
The second problem is far messier. It is the flood of AI-generated material aimed not at the machinery but at the voters themselves. Fake audio of a party leader saying something they never said, doctored video, swarms of automated accounts pushing a false claim in the final days of a campaign, or a convincing but entirely invented news story designed to suppress turnout in a particular electorate. None of that requires breaking into a single government computer. It targets what people believe, and it falls partly outside the control of any one agency. The Electoral Commission can secure its own systems, but it cannot police every social feed, and no minister can either.
That is where New Zealand’s answer starts to look thin compared with some partners. Canada set up a dedicated Security and Intelligence Threats to Elections taskforce back in 2019, pulling its intelligence and security agencies together to coordinate the defence of its elections in one place. RNZ asked Goldsmith whether he had considered a similar structure here. He did not answer the question. New Zealand’s approach remains spread across the Electoral Commission, the National Cyber Security Centre and other agencies, each with a piece of the puzzle but no single body holding the whole picture.
Whether that matters is a fair question for voters to weigh. A coordinated taskforce is not automatically better than a network of agencies talking to each other, and building new machinery in the months before an election carries its own risks. But the gap between asking for advice in July and having tested, working defences in place before November is real, and four months is not long in government terms.
The timing is the uncomfortable part. The warning from the Five Eyes agencies is not a forecast about some distant election cycle. It is about the tools that exist right now, in the hands of anyone with a laptop and a motive, during a campaign that has already begun in all but name. The advice Goldsmith has requested will tell the government how exposed the system is. What it does with that advice, and how fast, is the part that counts.
For voters, the practical takeaway is simple enough. Treat dramatic last-minute claims about candidates or the voting process with caution, check them against trusted sources before sharing, and know where the real information lives. If you are still working out where the parties actually stand before you vote, you can compare their positions using our voting tool.
What do you think? Is New Zealand doing enough to protect the November election from AI-driven disinformation, or do we need a dedicated taskforce like Canada’s? Share your view in the comments below.
This article was written by AI, briefed to report the facts, hopefully without some of the bias people bring to the job π
Leave a Reply