Jo Cribb explores the growing evidence that artificial intelligence is amplifying existing gender inequities, and why a more deliberate approach to AI governance is needed.

That AI has learnt what it knows from humanity, means it has the best and worst of us embedded at its core. Humans can be creative and empathetic but also come with hard-wired biases and the penchant for discrimination.

So, it not surprising that evidence of AI’s gendered harms is mounting.

AI-enabled violence against women is escalating, particularly for women in public life. Women are increasingly subject to automated harassment campaigns, explicit deepfakes, and sextortion. In New Zealand, one third of women report experiencing online abuse and the true figure is almost certainly higher.

Gender bias has been found in public sector AI systems that help allocate government services like welfare payments. Claude.ai had strong words about the framework that guides NZ public service use of AI: “The word ‘gender’ or ‘women’ does not appear in the Framework — a classic example of gender-neutral language masking gender-blind policy. The public service employs a disproportionately high number of women in the administrative and clerical roles most exposed to AI disruption. A governing framework that doesn’t mention gender is not a minor oversight; it is a significant policy failure.”

While many of us (me included) are enjoying AI’s productivity gains, there

is growing evidence of a gender adoption gap. Globally, women have consistently been found to be 20% less likely than men to use AI, and to be more cautious and less optimistic about its potential.

Part of the gap reflects unequal access: women are more concentrated in frontline caring roles and less represented in technology, both of which limit their exposure to AI tools and training.

But structural barriers are not the whole story. Research has found that when women do use AI in their work, they are seen as less competent than their male colleagues. Even when encouraged to use AI tools, women are penalised for “cheating”, their use of technology cited as evidence of inadequacy rather than initiative. Another study showed men are 27% more likely to be praised for using AI than their female colleagues.

Internet NZ found that women are less likely to say they understand AI and more likely to be concerned about it (51% of women compared with 42% of men). Not surprisingly, fewer women report using AI weekly: 69% versus 77% of men.

Sadly, addressing AI driven gender inequity and harm is not on the Government’s radar. Its 2025 Strategy for Artificial Intelligence according to Claude.ai “reads as though gender equality is someone else’s problem” and that “despite being a 19-page national strategy, the word ‘gender’ appears zero times in the main document and ‘women’ does not appear at all.” The AI concluded: “this is a significant omission given everything we know about AI’s disproportionate impact on women.”

But it is not too late.

Greater transparency about where AI is used in public services, and how those systems have been independently audited for bias, could provide a much needed boost in trust for public institutions.

We should follow other jurisdictions and introduce stronger regulation to prevent online harm. We should also rewrite the national AI strategy and the Public Service AI Framework through a gender lens to ensure we are harnessing the potential of AI for all and minimising its harm.

Creating such guardrails now can go some way to ensuring AI starts to reflect the best, not worst of us.

This draws on material previously published in a Newsroom column.

Jo Cribb, a previous Chief Executive of the Ministry for Women, now provides consultancy services on gender, AI, social policy and governance. She co-authored the recently published book Don’t worry about the Robots: How to survive and thrive in the new world of work.