8 ways to help your teams build lasting responsible AI (www.zdnet.com)

A new PwC survey of 310 executives finds that responsibility for “responsible AI” is shifting from centralized compliance teams to first-line builders — IT, engineering, data, and AI — with 56% saying those teams now lead governance. PwC frames this as a three-tier defense model (first line builds and operates responsibly; second line reviews and governs; third line assures and audits). The report also shows adoption stages (61% have integrated responsible AI into core ops, 21% are training, 18% are early-stage) and warns the big challenge is turning principles into scalable, repeatable processes. Experts underscore that unpredictable outputs from LLMs raise regulatory and operational risk, sometimes forcing organizations to re-scope or abandon projects. To help teams operationalize responsible AI, eight practical guidelines emerge: embed governance across the entire development lifecycle rather than bolting it on; use AI with clear, business-aligned purpose; set explicit policies and cross-functional steering committees; make oversight part of job roles and lifecycle governance (data sourcing, training, deployment, monitoring); keep humans in the loop; resist premature production pushes by mapping risks and checking model explainability; log and audit decisions on a 30–90 day review cadence; and tightly vet training data to avoid bias, IP exposure, and privacy/security gaps. Taken together, these steps aim to balance innovation with repeatability, trust, and regulatory resilience.