Rebuilding Trust: Queer in AI Approach to Artificial Intelligence Risk Management

Marginalized groups have historically been excluded from AI development and oversight, resulting in technologies that are less useful or actively harmful to these communities. Trustworthy AI requires intentional inclusion of these groups in design, deployment, and monitoring decisions, not just after-the-fact fixes.

Transparency, fairness, and accountability alone are insufficient for trustworthy AI. The paper argues that justice and shifting power to disempowered communities must be core values in AI systems, requiring a fundamental reframing of what "trustworthy" means beyond technical metrics.

Grassroots organizations like Queer in AI should be funded and empowered as essential partners in AI governance. These organizations bring expert knowledge about how AI affects specific communities and can provide credible, ongoing monitoring that external oversight bodies cannot.

Gender, sexuality, and queer identity require careful consideration in AI datasets and systems. Rather than simply excluding sensitive attributes, the paper advocates for thoughtful analysis of when and how these characteristics should be used, with specific mitigation strategies for potential harms.

Participatory design principles must be non-exploitative and feminist in nature. This means moving beyond tokenistic consultation to genuine power-sharing with marginalized communities throughout the entire AI lifecycle, supported by sustained funding and institutional commitment.