As generative artificial intelligence (AI) transitions from a novelty to a fundamental pillar of global economic and social infrastructure, a troubling trend has emerged: the technology is not merely mirroring human progress, but aggressively reinforcing the most regressive aspects of our history. From the drafting of corporate emails to the automated creation of marketing campaigns, AI systems are proving to be powerful conduits for discriminatory algorithms that disproportionately disadvantage women and marginalized communities.
Ahead of the UN Global Dialogue on Artificial Intelligence Governance and the AI for Good Global Summit in Geneva this July, UN Women has issued a clarion call to governments, developers, and corporate entities. The organization warns that without an immediate, radical shift in the design and governance of these systems, the technological future will be built upon the structural inequalities of the past.
The State of the Crisis: Pervasive Algorithmic Bias
The integration of AI into the modern workforce is nearly total. In the United Kingdom, for instance, 88 percent of advertising and media agencies have already incorporated generative AI into their daily workflows. However, the speed of adoption has vastly outpaced the implementation of ethical safeguards.
Recent research conducted on 133 distinct AI systems revealed a staggering prevalence of bias. Approximately 44 percent of the models tested demonstrated overt gender bias, while more than 25 percent exhibited a compounding effect of both gender and racial prejudice. These are not isolated instances; they are consistent, systemic patterns.
When large language models (LLMs) are tasked with completing sentences related to gender, the results are frequently distressing. Roughly one in five responses generated by these models lean into misogyny or sexism. The linguistic associations ingrained in these systems paint a clear, regressive picture: men are consistently linked with leadership, career advancement, and high-status industries, while women are persistently relegated to the domestic sphere, childcare, and familial support. In the most extreme cases, AI has generated content portraying women as sexual objects or even as property, stripping them of their agency in a digital reflection of the most archaic societal norms.
A Chronology of Neglect: From Training Data to Policy Gaps
To understand why AI is failing to uphold modern standards of equality, one must examine its foundation. These systems are not malfunctioning; they are performing exactly as they were designed, based on the vast, historical datasets upon which they were trained.
The Mirror Effect
For decades, the global archive of text—the primary source material for LLMs—was authored in a world defined by patriarchal structures. Women were historically "filed" under domesticity, while men occupied the corridors of power. As Jayathma Wickramanayake, UN Women Lead on Digital Technologies, noted in a recent interview, AI is simply extracting the biases embedded in the history of human communication.
The Policy Void
The crisis is exacerbated by a vacuum in governance. Of 138 countries surveyed by the UN, only 24 have explicitly mentioned gender within their national AI strategies, and a mere 18 have implemented substantive, gender-responsive measures. This suggests that the current state of bias is not an unfortunate "bug" or a temporary glitch; it is a policy choice. By failing to include gender-responsive frameworks in the early stages of development, regulators and tech leaders have effectively signaled that gender equity is not a priority in the digital age.
Supporting Data: The Cost of Inequality
The implications of these biases are multifaceted, affecting everything from professional security to personal safety.
The Erosion of Digital Safety
For women and girls, the risks are tangible. AI has lowered the barrier to entry for the creation and dissemination of online abuse. Data from UN Women indicates that nearly one in four women human rights defenders, journalists, and activists has experienced AI-assisted online violence.
The proliferation of deepfakes—manipulated images and videos created to harass or humiliate—has reached a point of crisis. Approximately 12 percent of women surveyed reported that their personal images were distributed without consent, while 6 percent have been direct targets of deepfake technology. As AI-generated content becomes indistinguishable from reality, the ability to track, prosecute, and prevent this form of gender-based violence is becoming exponentially more difficult.
The Economic Divide
The economic consequences are equally severe. The International Labour Organization (ILO) notes that women account for only 30 percent of the global AI workforce. This underrepresentation creates a "blind spot" in the development phase, as the people building these tools do not reflect the diversity of the populations that will eventually use them. Furthermore, women are nearly twice as likely as men to hold positions highly susceptible to automation. Without proactive intervention, the AI revolution threatens to widen the gender pay gap and push vulnerable, under-represented populations further into economic precarity.
Official Responses and the Call for Governance
The international community is reaching a tipping point. UN Women is advocating for a "human-rights-by-design" approach to AI. This means integrating gender-responsive measures at every stage of the AI lifecycle: from the curation of training data to the final deployment of commercial products.
The Business Case for Inclusion
Crucially, the argument for change is not merely ethical—it is commercial. The "Unstereotype Alliance," a UN-convened initiative, has produced research proving that inclusive advertising leads to superior business outcomes. Brands that eschew gender stereotypes in their marketing—often utilizing AI tools that prioritize diversity—see higher sales growth, increased customer loyalty, and stronger pricing power. Conversely, companies that rely on biased, stereotypical content risk severe reputational damage and long-term financial loss.
The launch of the Unstereotype Alliance playbook in June 2026 serves as a vital resource for marketers, providing a framework to identify and purge bias from AI-generated campaigns before they reach the public.
The Future: A Choice of Design
The path forward, according to experts, involves a fundamental re-evaluation of the relationship between technology and society. AI, if developed with intention and safety at its core, holds the potential to be a powerful tool for liberation. It could be used to detect and flag stereotypes, expand the reach of marginalized voices, and improve accessibility for those whom traditional systems have historically overlooked.
However, the realization of this potential depends entirely on who is sitting at the table. If the design rooms, policy-making bodies, and executive suites of major technology firms continue to exclude women, the technologies of tomorrow will inevitably solidify the inequalities of the past.
As stakeholders gather in Geneva this July, the directive from UN Women is unambiguous: gender equality must be a non-negotiable requirement of the AI architecture. We are at a juncture where we must decide whether we want to build a digital future that elevates humanity or one that traps it in the prejudices of previous centuries. The data is clear, the risks are documented, and the solution requires a collective, global commitment to ensuring that the next generation of intelligence is truly representative of the entire human experience.
Asep Darmawan
rifanmuazin
Iffa Jayyana
