US Seeks Global Dominance in Open-Source AI Development

The US government has set its sights on making American open-source AI models the global standard, with National Cyber Director Sean Cairncross announcing a vision for 'preferential adoption by planet Earth' at the Black Hat conference in Las Vegas.

Cairncross emphasized that this goal would be achieved through competitive superiority rather than mandates, framing open-source development as an innovation engine particularly beneficial to startups and smaller firms.

The policy direction is connected to a broader AI executive order signed in June 2026, which prioritizes voluntary security testing and information sharing over heavy-handed regulation.

However, the government has carved out open-weight models from its security framework, creating an awkward gap. Open-weight models are those where the architecture and trained parameters are publicly available but the training data and methodology might not be.