Webinar | Session 4: Ethics, Bias, Deepfakes and Identity Harms

Value of the Information & Strength of the Argument

The webinar provides valuable insights into the regulatory landscape for AI in South Africa, particularly regarding child protection. Eugene Mashaba offers concrete data on AI-generated CSAM, including a specific incident with Grok AI, and discusses the Film and Publications Board’s role. Marcus Angula builds a strong legal argument, referencing key acts and international conventions, and effectively argues that enforcement is the primary challenge. The argumentation is coherent and grounded in legal expertise, though some claims could benefit from more direct citations.

Scientific Rigor, Source Quality, Title Accuracy

The speakers reference credible sources such as UNICEF and the Film and Publications Board, and cite specific South African legislation. However, some statistics are mentioned without direct links, and the webinar lacks a formal bibliography. The title accurately reflects the content, which is focused on the stated themes. The discussion is rigorous in its legal analysis, but the absence of detailed source citations in the presentation limits its scientific rigor.

The title accurately reflects the content, which focuses on ethics, bias, deepfakes, and identity harms in the context of South African AI policy.

7/10

The webinar features expert speakers from regulatory and legal backgrounds, providing specific data points and referencing legal frameworks. However, some claims lack direct citations, and the discussion is largely opinion-based.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction and welcome by the moderator, outlining the session's focus on ethics, bias, deepfakes, and identity harms.
• 02:00Eugene Mashaba begins his presentation on child protection and AI governance, citing UNICEF data on AI-generated CSAM.
• 05:00Mashaba discusses the Grok AI incident, where over 23,000 sexualized images of children were created in 11 days.
• 10:00Mashaba highlights the 1,200% rise in deepfake-related fraud in South Africa and the need for a centralized AI policy.
• 15:00Mashaba explains the Film and Publications Board's regulatory role and the importance of collaboration through the ICT Regulators Forum.
• 20:00Mashaba emphasizes the need for proactive awareness and grassroots education on digital safety.
• 25:00Marcus Angula begins his presentation, focusing on legal frameworks for protecting children from synthetic images.
• 30:00Angula discusses the Cyber Crimes Act, Children's Act, POPIA, and Sexual Offences Act, and their relevance to AI-generated content.
• 35:00Angula argues that enforcement is the main challenge, and that notice-and-takedown mechanisms are insufficient.
• 40:00Angula concludes by emphasizing the need for platform accountability and a rights-based approach to AI governance.

• UNICEF report on AI-generated CSAM — Cited by Eugene Mashaba regarding the impact of AI-generated child sexual abuse material on 1.2 million children in 11 countries.
• Grok AI incident report — Cited by Eugene Mashaba regarding the creation of over 23,000 sexualized images of children in 11 days.
• Film and Publications Act — Mentioned by Eugene Mashaba as the legal basis for the Film and Publications Board's regulatory powers.
• Cyber Crimes Act — Mentioned by Marcus Angula as a legal framework criminalizing non-consensual sharing of intimate images.
• Children's Act 2005 — Mentioned by Marcus Angula as establishing child rights to protection from harm and exploitation.
• Protection of Personal Information Act (POPIA) — Mentioned by Marcus Angula as regulating the processing of biometric data, including facial images.
• Sexual Offences Act 2007 — Mentioned by Marcus Angula as criminalizing child sexual abuse material and exploitation.
• Convention on the Rights of the Child — Mentioned by Marcus Angula as the international human rights framework for child protection.

• UNICEF report on AI-generated CSAM — Supports the claim of widespread AI-generated child sexual abuse material.
• Grok AI incident report — Illustrates the scale of AI-generated harmful content.