Four govt officials suspended for fake AI policy citations

In an embarrassing climbdown, the Department of Communications and Digital Technologies (DCDT) has announced the suspension of two officials with immediate effect, following the withdrawal of its long-awaited draft AI policy – for using academic citations widely believed to be AI hallucinations.

“The draft policy was recently withdrawn from public consultation following revelations that fictitious references were used during the policy development process. The irresponsible use of AI tools compromised the integrity of the policy document,” the department said in a media statement.

Read:

In the same week, the Department of Home Affairs (DHA) announced it was suspending two senior officials after detecting apparent AI “hallucinations” cited as references in the cabinet-approved Revised White Paper on Citizenship, Immigration and Refugee Protection.

Disciplinary and internal processes

The DHA has appointed two independent law firms to manage the disciplinary process against the suspended officials and to review all policy documents produced by the department dating back to 30 November 2022, when AI tools first became widely accessible to the public.

In future, the DHA said it will design and implement AI checks and declarations as part of its internal approval processes.

In a statement issued last week, the department said the fake references appear to have been appended to the policy document after the fact, although “the body of the Revised White Paper continues to accurately reflect the government’s position on the policy reforms required in these domains.

“The Department nonetheless sincerely apologises for this unacceptable oversight. The immediate implementation of precautionary suspensions, independent review processes, as well as internal reforms reflects our commitment to taking any and all steps required to rectify the situation and avoid any repeats in future.”

This has prompted calls for stricter controls over the use of AI in government, with questions being asked whether similar hallucinations may have affected other policies already in place.

Home Affairs Minister Leon Schreiber said all departments occupied by Democratic Alliance members will be required to implement AI verification as part of the policy document approval process.

“In addition, I will raise at the next meeting of the Cabinet the urgent need for this approach to be implemented across government. The development of AI presents extraordinary opportunities, but it must also be used responsibly and with integrity. We will ensure that is the approach into the future,” said Schreiber in a statement.

Policy flaws uncovered

News24 first reported that at least six identifiable references cited in the DCDT’s draft AI policy did not exist.

This prompted the department to withdraw the policy document and commence an internal review, leading to the suspension of the two officials.

“This matter remains under investigation, and the outcome will be communicated in line with the Department’s commitment to transparency on the matter,” the statement added.

Listen/read:

Public Interest SA commended Communications and Digital Technologies Minister Solly Malatsi and the DCDT for the swift withdrawal of the flawed draft and for initiating an internal review.

“The minister didn’t obfuscate, he didn’t try blame shifting, he took responsibility and asked for heads to roll, which is what we expect of all ministers,” Public Interest SA chair Tebogo Khaas told the SABC.

Incidents such as these will undermine public trust in government policy-making and highlight the need for greater vigilance among officials, including cabinet members.

It hasn’t gone unnoticed that the two departments involved are headed by ministers from the DA, raising questions about whether misuse of AI may be more widespread than reported but less visible in departments led by other parties.

The question now is how to prevent such incidents in future.

Khaas believes penalties must be commensurate with the transgression, but more importantly, robust mechanisms must be embedded within government to detect and prevent AI hallucinations from creeping into policy development.