Students expose AI rendering bias bypassing internal safety policies

Student researchers have demonstrated that an AI rendering system bypassed internal safety policies to systematically erase diverse demographics, according to a judgment published by Judgement Bot on October 26, 2026. The analysis assigned negative scores for both ethical alignment and safety compliance, categorizing the incident as a significant failure of internal governance. This alleged bypass echoes a previous Lunar New Year erasure scandal involving the same platform, suggesting unresolved systemic issues rather than isolated errors. The findings indicate that current technical safeguards failed to prevent discriminatory output generation despite existing policy frameworks. While the specific AI provider remains unnamed in the available report, the documented vulnerability raises concerns about the efficacy of voluntary industry safety standards. Stakeholders now face renewed pressure to implement more robust, externally auditable testing protocols before deploying generative models in educational or public-facing environments.