As artificial intelligence (AI) is rapidly reshaping digital spaces, the harms it can enable and amplify are evolving just as quickly. However, those who are most vulnerable and bear the greatest impact – namely women, queer, gender-diverse people and marginalised communities – are the ones most often left out of policies governing the development and use of AI. In addressing technology-facilitated gender-based violence (TFGBV), governance frameworks are struggling to keep pace.
This was the central concern explored at the launch of APC’s new research report, Bridging the gap: Addressing technology-facilitated gender-based violence in global AI governance, held on 12 August. Moderated by Verónica Ferrari, APC’s global policy advocacy lead, the launch brought together report author Daniela Schnidrig, a digital rights and policy consultantspecialist, as well as Rebecca Ryakitimbo, a feminist technologist, researcher and community leader, and Marina Meira, public policy coordinator at Derechos Digitales.
Together the panel discussed key learnings from the report, which examines how AI governance frameworks address gendered risks and how existing TFGBV norms respond to harms introduced or amplified by AI.
The findings point to a significant gap that is central to the report: AI frameworks and TFGBV norms are being developed? largely in parallel, sometimes even by the same multitlateral institutions, with limited connection between the two agendas. As Schnidrig explained during the launch, “There are now these great many AI governance frameworks and this growing body of international and regional norms on TFGBV [but they] rarely or barely speak to each other... Who bears the cost of that gap and how do we break it?” Meanwhile, among the most urgent and under-regulated consequences of AI deployment is the rise of automated harassment, deepfakes, gendered disinformation and surveillance-enabled abuse.
AI is changing the scale of gendered harms
The research maps three intersectional areas: structural gendered risks in AI; TFGBV harms enabled or amplified by AI; and governance gaps and challenges. These include the gender digital divide, data extraction and lack of consent, gendered disinformation, and bias and stereotypes, among other emerging topics.
The reality, as each of the speakers noted, is that AI does not create entirely new forms of violence; it can also dramatically increase the scale, speed and persistence of existing harms. “Coordinated harassment campaigns or gender disinformation campaigns before maybe would have required an organised human effort that now can be run at volumes and with a persistence that was simply either not possible before or very costly,” Schnidrig noted.
The consequences extend beyond individual experiences of abuse. The report highlights the economic and material impacts of TFGBV, including people leaving their jobs, withdrawing from public life and relocating because of threats to their reputation. AI-enabled TFGBV can create a chilling effect that, according to Schnidrig, “leads to women, queer, gender-diverse people being absent from these civic spaces, academic spaces, policy discourse, because the cost of their presence is too high.”
This absence is itself a structural harm, restricting freedom of expression and political participation, with the detrimental consequence of reinforcing existing inequalities. “Voices that are missing from the public debate then end up being these same voices that are missing from the rooms where AI is being governed,” Schnidrig observed.
Gender cannot be an add-on
As a case study, the participants discussed the recent UN Global Dialogue on AI Governance and what it revealed about the place of gender and intersectionality in global governance processes.
For Ryakitimbo, gender remained largely peripheral to the dialogue. “The conversation around gender was something that stakeholders took upon themselves to mention it, but it was not mainstreamed,” she said. TFGBV, she added, was frequently absorbed into broader conversations about AI security rather than treated as a specific concern requiring dedicated attention.
Intersectionality was also largely absent, and as the speakers noted, gender cannot be understood in isolation from race, age, disability, geography, class and other dimensions of inequality. Without this lens, governance risks overlooking how different communities experience AI-enabled harms.
Meira similarly stressed that representation and mainstreaming are not the same thing. She explained that having women participate in AI governance spaces matters, but “a gender-balanced panel does not necessarily produce a gender-responsive discussion.”
Mainstreaming is the most effective way to ensure policies match realities. As Meira explained, “gender mainstreaming is asking – across every dimension of AI governance – who benefits, who bears the risks, whose labour and resources sustain this A system, whose knowledge is represented, who participates in decisions, and who has access to remedy when harm is made.”
From principles to implementation
A critical element noted by the speakers in discussing the research is that most AI frameworks barely consider TFGBV. Where gender is addressed, it is often discussed at the level of broad principles such as fairness, diversity and non-discrimination, rather than through concrete measures that create obligations. Meanwhile, many TFGBV instruments predate current AI technologies and have not been sufficiently updated to address things like deepfakes, algorithmic amplification or other AI-enabled forms of abuse that are becoming more rampant.
Schnidrig emphasised that this distinction between principles and implementation is crucial. “It’s comparatively easy… to get gender equality into a preamble,” she said during the discussion. The challenge, she explained, is ensuring that gender considerations “survive the journey into the operative provisions.” This can be done through assessments, audits, disaggregated data requirements, budget allocations and implementation plans.
Bridging the silos
So how do we bridge these glaring gaps? The speakers had some concrete suggestions, grounded in the extensive research conducted. They echoed recommendations within the report, which proposes action across three areas: AI-specific measures, TFGBV-specific reforms and crosscutting structural changes. These include gender-responsive risk and human rights impact assessments throughout the AI lifecycle; explicit recognition of TFGBV as a category of AI risk; updated definitions of digital violence; stronger reporting and redress mechanisms; intersectional data collection; regulatory capacity-building; and meaningful participation of women, gender-diverse people and civil society.
While governments have an important role in adopting and enforcing effective legal, policy and regulatory measures to ensure that the private sector actors – particularly large technology platforms, infrastructure providers and AI developers – identify, prevent, mitigate and account for TFGBV risks, multilateral bodies can support this through norm setting and coordination, while civil society can play a key role in monitoring and documenting impacts.
Breaking down institutional silos is at the heart of these recommendations. “We need to find sort of a mechanism for… different instruments to be able to communicate to each other,” Ryakitimbo proposed. These include everything from gender policies and ministries to AI strategies and governance mechanisms.
Meira also pointed to the value of strengthening existing human rights mechanisms rather than treating AI governance as an isolated field. Since AI cuts across labour, environmental, data, gender and human rights governance, these existing institutions need the capacity and mandates to address its implications.
Ultimately, bridging the gap calls for a shift away from reactive, fragmented regulation towards integrated, rights-based governance. The first challenge is to acknowledge that AI can produce and is largely producing gendered harms, but that is not the end of the story. It is imperative to ensure that gender equality and intersectionality shape how AI is designed, developed, deployed and governed. To this end, the report calls on governments, international institutions and private sector actors to ensure inclusive, equitable and sustained participation of women, gender diverse communities and civil society actors across all stages of the AI lifecycle.
As the speakers made clear, the opportunity to bridge these agendas is now. If AI governance continues without gender and TFGBV at its core, the people most affected by its harms risk remaining absent from the public spaces where those harms are felt and the policy spaces where the rules are made.