QCon London 2026: Ethical AI Is an Engineering Problem

Clara Higuera's framing at QCon London 2026 hits at something we've been wrestling with in production: ethical AI risks aren't solved by compliance checkboxes. They're baked into data pipelines, model training loops, and deployment automation. When I look at our data platform, the real issues are architectural—how we version datasets, track lineage for bias auditing, implement feature stores that prevent leakage, and structure CI/CD pipelines to catch fairness regressions. This reframes ethical AI as a core engineering discipline, not a bolt-on governance layer. For teams building modern data stacks, this means treating responsible AI like you'd treat reliability or security: embedding it into infrastructure decisions from day one. That means investing in data quality frameworks, feature governance tooling, and observability systems that surface distributional shifts. The competitive advantage goes to organizations that make ethical considerations inseparable from their data architecture, not those treating it as an afterthought.