Adji Bousso Dieng argues AI safety fears can entrench tech monopolies

AI policy should confront harms already affecting people, not let warnings about future catastrophe become a reason to concentrate power, argues researcher Adji Bousso Dieng. In an Al Jazeera opinion piece published on 17 October, she makes the case for preserving open-source AI while taking present-day risks seriously.

Watch Desk analysis

What happened

Dieng argues that calls to restrict open-source models, made in the name of safety, could protect the position of a few large AI companies and leave other countries dependent on their systems. She says open-source tools can help local developers adapt AI to their own needs, especially where building frontier models from scratch is too expensive.

She points to Kenya’s Social Health Authority system as an example of the stakes: Dieng says an imported AI system misjudged incomes in the country’s informal economy, leading to excessive health-premium charges for the poorest people. Her argument is that local adaptation matters, not that open source alone guarantees a better system.

Why it matters

This is a challenge to the idea that safety and openness are opposing choices. Dieng calls for robust safety research and testing, while warning that rules shaped around hypothetical future risks could restrict competition and local control today. Her examples, and her broader claims about industry motives, are the author’s argument rather than independently established findings.

Our read

The useful question is not whether AI should be safe or open. It is who gets to define safety, who can still build and adapt systems, and whether the rules address demonstrated harms as well as feared ones. The Kenya example makes that more than an abstract argument about model access.

What to watch

• Whether AI rules distinguish among model types and levels of risk.
• How local developers can adapt systems while testing and addressing real harms.
• Whether policy debates give as much attention to current impacts as to future risks.

Discussion spark: Should AI safety rules prioritise limiting access to powerful models, or preserve broad access and focus on testing how systems are used in practice?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.