“Is DEI a dirty word for AI?” - Check Point’s responsible AI warning (www.techradar.com)

Check Point’s enterprise team, via Charlotte Wilson at the Cyber Leader Summit, warned that generative AI systems—trained on vast swaths of scraped internet data (articles, social posts, videos, books)—inherit and amplify human bias. Because many consumer models optimize for engagement or “what the user wants” rather than strict accuracy, they can be sycophantic, hallucinate, and repeat discriminatory patterns. That danger is now business-critical: companies use these models for hiring, HR and decision-making, and have already faced legal exposure (e.g., allegations against Workday over age discrimination). Wilson argues the internet’s polluted data—exacerbated by adversarial content—means bias can’t be fully eliminated at the source. The practical response she outlines is operational governance: new roles for “AI checkers” to spot-check outputs for safety and fairness, board-level oversight that explicitly includes human-fairness reviewers, and rigorous testing, fact-checking and repeated validation. She also warns that political rollbacks of DEI initiatives reduce organizational appetite to proactively correct inequities, raising the likelihood that AI will perpetuate them unless firms are purposeful about deployment. For AI/ML teams this means prioritizing provenance-aware datasets, human-in-the-loop pipelines, bias-monitoring tooling, and clear accountability to limit legal and ethical risk.