What Data Can’t See: Building Queer Safety Into the First Line of Code

A product can work exactly as designed and still make someone unsafe. By the time a report button or block feature gets built, teams have already made hundreds of decisions about who their users are, which identities exist, what data matters, and what the system is allowed to infer. For queer communities, those assumptions have real consequences.

This is a working session, not a keynote. Drawing on experience in data science, product leadership, and building She & HER for the Sapphic community, we'll look upstream of "AI bias" at the human and product decisions that shape technology before a model produces a single output. You'll leave with a five-question safety framework and a way to apply it from personas and research through data, design, implementation, and iteration:

1. Who could this feature make less safe?
2. What are we assuming about the person using it?
3. What does our data not know?
4. What happens when our system is wrong?
5. How does the user regain control?

We'll run the framework live against a real product feature, then safety-test attendee examples together.

Who it's for: founders, product leaders, designers, researchers, data scientists, ML practitioners, engineers, and trust, safety, privacy, and community teams. No technical background required.

Bring: a laptop or tablet and one real feature, workflow, dataset, AI use case, or product decision you're working through. Nothing formal, no code, no proprietary details.

Interactive session. Laptop encouraged. Bring a feature. Leave with a safety audit.