South Africa withdraws AI draft policy after AI‑generated fake references exposed

TL;DR: South Africa retracted its national Artificial Intelligence (AI) policy draft after discovering its reference list contained fabricated, AI-generated sources. This withdrawal highlights the critical need for vigilant human oversight in AI-assisted policy development.

South Africa's Department of Communications and Digital Technologies published its draft national Artificial Intelligence (AI) policy on April 10. The government sought public and stakeholder feedback on this crucial document, which aimed to establish a framework for regulating AI use within the country. The policy framework intended to promote an ethical, inclusive, and human-centered governance of AI, ensuring alignment with South Africa's constitution and national development priorities. It served as a foundational roadmap for navigating the rapidly evolving technological landscape.

Minister Solomon Malatsi confirmed the presence of fictitious sources within the draft AI policy's reference list. This revelation followed internal inquiries initiated after public scrutiny. The minister explicitly stated that the draft contained various fabricated sources, acknowledging that AI-generated citations were likely included without proper verification. This critical error compromised the document's integrity and credibility, leading to the immediate withdrawal of the entire draft national AI policy.

This incident underscores the inherent risks of integrating artificial intelligence tools into critical government functions without stringent human verification. The failure to identify fabricated information in a national policy document raises significant concerns about the drafting processes and quality assurance protocols within governmental departments. Minister Malatsi emphasized that South Africans deserve a higher standard, stating that the department did not deliver on acceptable standards for an institution leading digital policy. He also highlighted the necessity of vigilant human oversight when utilizing AI, a lesson he stated the government takes with humility. Future developments will likely include revised drafting procedures, enhanced verification safeguards, and accountability measures for those involved. Watch for how South Africa re-establishes trust in its digital policy framework and implements these new protocols.