Bias in AI isn’t just a bug—it’s a reflection of the world we live in. In this TFiR clip, Jesse McCrosky, Principal Architect – GenAI at Egen, unpacks what fairness in AI really means, and how governance provides a structured path to responsible outcomes.

Bias Is Baked In

McCrosky begins by debunking the myth that data can be made “fair” or models made “unbiased.” From healthcare inequities in data generation to gaps in collection and curation, bias is everywhere—and that includes model architecture.

“There’s no such thing as unbiased data,” he says. “And there’s no clear answer to what fairness should look like.”

Governance as Risk Mitigation

So what can we do? McCrosky offers a clear answer: AI governance isn’t about perfection. It’s about mitigation.

By treating fairness as a risk management challenge—not a binary outcome—organizations can identify harms early, track decision logic, and iterate responsibly. He references OpenAI’s DALL·E, which once returned only white male lawyer images. Their fix (though imperfect) reflected a governance mindset: identify risks, act transparently, and iterate.

Toward a Better AI Culture

Ultimately, McCrosky urges organizations to shift the conversation. It’s not about building perfect models—it’s about building processes and cultures that understand trade-offs, document decisions, and care enough to ask hard questions.

Why AI Governance Is the Innovation Catalyst Every Organization Needs