Biasis a common risk acrossboth proprietary and open-source models, andnot uniqueto proprietary deployments. All AI systems - regardless of origin - require evaluation for fairness, accuracy, and representativeness.
From theAI Governance in Practice Report2025:
"Bias, discrimination and fairness challenges are present in both open and closed models, regardless of how the model is sourced." (p. 41)