The small African lab sounding the alarm on AI bioweapons

Rest of World
Read full story· 8 min
In 2024, MinoHealth AI Labs in Ghana removed safety limits from their drug discovery AI and generated thousands of potentially toxic molecules, highlighting risks of AI-enabled bioweapons. This study has influenced global discussions on DNA synthesis regulation and biosecurity, with figures like Bill Gates acknowledging the threat.

What happened

In 2024, researchers at MinoHealth AI Labs in Accra, Ghana, intentionally disabled the safety features on their drug discovery AI system to see if it could design toxic substances. The AI generated over a thousand protein designs and five thousand small molecules, all predicted by toxicity software to be harmful, with some resembling known poisons like ricin. The team did not synthesize any of these substances, but published their findings on arXiv in May 2025. Their work has since been referenced in international discussions about biosecurity and the regulation of DNA synthesis providers.

Key details

  • The AI model generated 1,020 protein designs and 5,000 small molecules.
  • Toxicity prediction software rated all generated substances as likely poisonous.
  • Some AI-generated molecules were structurally similar to ricin.
  • The study was posted on arXiv in May 2025.
  • Bill Gates publicly warned about AI-enabled bioweapons in September 2025.

Why it matters

This event has brought new urgency to governments, DNA synthesis companies, and biosecurity organizations, who now face the challenge of preventing misuse of AI tools for creating dangerous biological agents. The study has prompted high-level discussions about regulating DNA synthesis and has been cited by international agencies and public figures. For technology companies and researchers, there is now increased scrutiny and pressure to strengthen AI safety measures and oversight.


More in Public Sector