Key Takeaways
- AI Evidence in Courts: US cases use AI prompts as proof; South Africa has no reported AI prompt evidence yet, courts slow to adapt.
- Privilege Risks: Using public AI tools risks losing attorney-client privilege; secure, firm-approved AI tools recommended to protect data.
- AI Hallucinations: Inaccurate AI citations cause warnings but weak enforcement; courts may soon demand AI prompt disclosure for challenged evidence.
- Regulatory Gaps: Law Practice Council lacks enforcement on AI ethics, creating a “Wild West” environment in legal AI use.
- AI Tool Use: Paid AI versions offer better privacy; staff must be trained on secure use to avoid accidental data leaks and privilege loss.
- Cultural Challenges: Inconsistent AI understanding and poor ethics worsen AI risks; firms need clearer policies, monitoring, and cultural shifts toward cautious use.
Notes
AI Usage in Legal Evidence
The discussion centred on emerging court cases where AI-generated prompts and transcripts have been used as evidence, highlighting risks and evolving legal implications.
- US Court Cases and AI Evidence as Proof (06:25)
- Conservation Law Foundation vs. Shell: AI prompting was submitted to prove an expert’s methodology and authenticate testimony.
- US vs. Heppna: Involved chatbot transcripts obtained by prosecutors, bringing attorney-client privilege issues to light.
- South African Legal Context and Adoption (05:50)
- AI prompts have not yet been formally admitted or widely introduced as proof in South African courts, reflecting scepticism regarding the judiciary’s readiness.
- Adoption and concern are currently slowed by a lack of regulatory enforcement and specialized AI expertise within the judiciary, though relevance is expected to grow.
- Internal Law Firm Risks and Confidentiality (09:51)
- Uncontrolled use of free AI tools externally by article clerks or junior staff risks breaching confidentiality, voiding privilege, and potentially exposing data to subpoenas or court exposure.
- Mitigation: Malcolm urged transitioning to paid versions within secure, firm-approved infrastructure (such as Gemini’s pro tools, which do not use data for training).
- AI Hallucinations and Professional Ethics (20:48)
- South African lawyers have faced reprimands for relying on AI-hallucinated cases in court, threatening case integrity and professional ethics amid broader governance challenges and limited Law Practice Council (LPC) enforcement.
- Future Outlook: Malcolm predicted that if hallucinated evidence is challenged, courts will increasingly demand to inspect the underlying AI prompts.
And a little humour shared by my wife…..
