AIBRIEF
At the New Delhi summit, Ramaphosa outlined four principles for an international AI evaluation mechanism, pairing safety style safeguards with a developing economy capacity ask.
South African President Cyril Ramaphosa used his slot at the 18th BRICS Summit in New Delhi on 13 September 2026 to pitch the bloc's peers on an international, independent scientific body that would evaluate AI models, and to weld a "sovereign AI" capacity track for developing economies onto the same proposal.
The four principles, per SAnews and Xinhua's on-scene dispatch: meaningful human control over AI systems; mandatory reporting of serious incidents; safeguards that grow stronger as models become more capable; and investment in domestic AI research and compute for developing countries. "South Africa accordingly proposes that BRICS should establish an international mechanism for the independent scientific evaluation of AI," Ramaphosa said.
The pitch lands on day one of a summit whose closing declaration is not yet public. Whether China, India, Russia, and Brazil sign on is the open question. Tribune's framing of the move as echoing Anthropic CEO Dario Amodei's same-day call for external oversight is an editorial linkage; Ramaphosa's quoted remarks do not name Amodei.