\n<\/p>\n
South Africa\u2019s Department of Home Affairs has suspended two senior officials with immediate effect after AI-generated hallucinations were discovered in the reference list of its revised white paper on citizenship, immigration and refugee protection \u2014 the second cabinet-approved policy document caught with fabricated sources in less than a week.<\/p>\n
A chief director in the unit responsible for the white paper was placed on precautionary suspension Thursday afternoon, with a director involved in the drafting process to follow Monday, the department said in a statement. Home Affairs has appointed two independent law firms \u2014 one to manage the disciplinary process and another to review every policy document the department has produced since Nov. 30, 2022, the day OpenAI\u2019s ChatGPT was released to the public.<\/p>\n
The department also committed to designing AI checks and declarations into its internal approval processes going forward.<\/p>\n
The revised white paper, championed by Home Affairs Minister Leon Schreiber as the most fundamental reform of South Africa\u2019s citizenship and immigration framework in a generation, was approved by Cabinet on April 3. An initial review by the department suggests the suspect references were generated and added to the document after the fact and are not cited in the body of the text. The reference list has been withdrawn pending the outcome of the independent probe.<\/p>\n
Home Affairs maintained that the body of the white paper \u201ccontinues to accurately reflect the government\u2019s position\u201d and is \u201cnot materially affected\u201d by the apparent hallucinations. \u201cThe department nonetheless sincerely apologises for this unacceptable oversight,\u201d it said, describing the episode as \u201can opportunity to further modernise our internal processes.\u201d<\/p>\n
The suspensions come four days after Communications and Digital Technologies Minister Solly Malatsi withdrew the entire Draft National AI Policy after a News24 report revealed that at least six of its 67 academic citations were fabricated \u2014 fictional articles attributed to real journals and to authors who had never written on the topics in question. That policy was approved by Cabinet on March 25 and gazetted for public comment on April 10.<\/p>\n
Malatsi said in a statement that the most plausible explanation was that \u201cAI-generated citations were included without proper verification,\u201d pledging \u201cconsequence management\u201d for those responsible for drafting and quality assurance. He called the episode \u201can unacceptable lapse\u201d that demonstrated why \u201cvigilant human oversight over the use of artificial intelligence is critical.\u201d<\/p>\n
Khusela Diko, chair of the parliamentary portfolio committee overseeing Malatsi\u2019s department, suggested the ministry \u201cskip using ChatGPT this time\u201d when redrafting the AI policy.<\/p>\n
The back-to-back incidents raise pointed questions about how widely AI use without verification has taken hold across South Africa\u2019s public sector \u2014 presumably the reason Home Affairs has committed to reviewing every policy document produced over the past three and a half years. The episodes also raise broader questions about the quality assurance processes that allowed two cabinet-approved documents in quick succession to clear internal scrutiny with manufactured references attached.<\/p>\n<\/div>\n
\n