Global Regulators Demand AI Safety Overhaul to Protect Women from Deepfakes

The rapid democratization of generative AI tools has lowered the barrier to entry for malicious actors, transforming what was once a niche technical skill into a widespread weapon of digital harassment. In a significant move toward global AI governance, Hong Kong’s Office of the Privacy Commissioner for Personal Data (PCPD) recently co-signed a joint statement with over 60 international organizations. This coalition aims to address the escalating threat of technology-facilitated violence (TFV), specifically targeting the misuse of deepfakes. The core of their argument is that while digital abuse is not a new phenomenon, artificial intelligence has fundamentally altered its speed, scale, and sophistication, creating an environment where women and girls are disproportionately victimized.

Statistical evidence underscores the gendered nature of this technological crisis. Current estimates suggest that approximately 90 percent of non-consensual deepfake pornography depicts women. This data points to a systemic failure in how AI models are currently trained, deployed, and monitored. The transition from traditional forms of harassment—such as telephone stalking and doxxing—to AI-generated deepfakes represents a shift from manual abuse to automated, high-fidelity character assassination. For the victims, the impact is often permanent and devastating, yet the institutional response from technology platforms remains largely inadequate. The PCPD and its global partners argue that the current 'reactive' model of safety—where content is reported and removed only after it has spread—is fundamentally broken.

“

The transition from traditional forms of harassment—such as telephone stalking and doxxing—to AI-generated deepfakes represents a shift from manual abuse to automated, high-fidelity character assassination.

Industry experts and regulators are now calling for a 'safety-by-design' approach. This philosophy suggests that the safety of vulnerable populations should be integrated into the very architecture of AI models and social media platforms. Currently, many platforms suffer from a 'transparency gap,' where reporting mechanisms are opaque and the follow-up process is slow or non-existent. For many women, navigating these reporting systems is a retraumatizing experience that reinforces a sense of exclusion from digital spaces. By placing women and girls at the heart of the technology lifecycle—from initial dataset curation to final governance frameworks—developers can begin to identify and mitigate biases before they manifest as harmful outputs.