Rest of World · Technology

Africa Demands Voice in Global AI Safety Standards

African nations are pushing for a greater role in setting international AI safety standards, concerned about the deployment of foreign AI systems without adequate local safety evaluations.

African leaders at the UN called for co-architecting AI standards, as foreign AI models are piloted across the continent with insufficient context-specific safety checks.

Experts note a vacuum in AI safety discussions in Africa. Over 80% of surveyed large companies are running AI pilots, often without formal safety oversight.

AI misuse incidents are emerging, including AI-driven health insurance cost hikes in Kenya and AI-powered blackmail schemes in West Africa.

The key AI model distinction is control ('off switch'), not origin. Open-weight models are harder to shut down remotely than closed-weight ones.

Despite AI use in sensitive sectors, fewer than half of African countries have AI policies. Kenya is in an international AI safety network, but its focus is global, not African-specific.

There's a need for 'Africa-centric' safety evaluations, assessing harm by country's economy. Third-party, region-specific risk assessments are crucial.

African nations seek to build capacity for independent evaluations, aiming to be governing partners, not just markets or testing grounds for AI.

AI-samenvatting op basis van de bron.