New Zealand criminalises sexualised deepfakes, law advances

New Zealand is advancing legislation to criminalise sexualised deepfakes, with a bill scheduled for its first reading this week, The Conversation reports. The Conversation states the proposed amendment would make creating, sharing or selling sexually explicit deepfakes without consent a criminal offence. Reporting by The Conversation, Stuff and RNZ documents that Grok, the AI chatbot on X, has been used to produce sexualised images of real people without their consent.

Stuff reported that Minister of Internal Affairs Brooke van Velden has tasked officials to provide a report and advice before further steps are taken. RNZ quoted ECPAT national director Eleanor Parkes calling the phenomenon "hugely problematic" and saying "we can't rely on goodwill" in addressing AI-enabled creation of sexual images.

Industry-pattern observations: Criminalisation addresses the actor and the content after the fact, but comparable regulatory landscapes show enforcement alone often struggles to keep pace with rapid model iteration and distribution channels. Platform-integrated generative models such as Grok combine large-scale user reach with ease of prompting, which amplifies the speed at which non-consensual images can spread across social networks and repositories. For practitioners, this pattern implies that legal deterrents will likely need to be complemented by engineering controls, content provenance mechanisms, and robust moderation tooling to materially reduce production and sharing at scale.

Industry-pattern observations: Other jurisdictions have recently moved to criminalise or regulate non-consensual deepfakes and to pressure platforms to restrict generation and distribution. Reporting on X and Grok in early 2026 prompted regulatory scrutiny in multiple countries, creating a template where national laws and platform enforcement interact. For AI safety teams and compliance officers, those interactions are a practical area to anticipate: legal changes tend to raise the bar for platform policies, takedown processes, and provenance or watermarking standards.

• •Legislative milestones: whether the bill passes subsequent readings and the exact statutory language defining consent and sexualised content, as initially reported by The Conversation.
• •Executive guidance and regulator action: whether officials asked for by Minister van Velden (Stuff) recommend platform-specific restrictions, age limits, or a dedicated regulator, as queried in public reporting.
• •Platform responses and technical mitigations: whether X, AI model providers, and third-party app stores deploy prompt restrictions, filtering controls, or authenticated image provenance to limit non-consensual generation and downstream sharing.

Industry-pattern observations: Product teams and ML ops groups should expect increased demand for explainability around training data provenance, improved safety-by-design controls for image generators, and operational tooling that integrates legal takedown workflows with detection signals. Privacy, trust and safety, and legal-compliance teams will be the primary internal consumers of those capabilities.

The reporting assembled from The Conversation, Stuff and RNZ documents both the legislative move in New Zealand and the operational problem space exemplified by Grok. RNZ and advocacy groups emphasise the limits of voluntary platform action, while The Conversation frames criminalisation as an important but incomplete response.