http://x.com/i/article/2050188347834847232
Reflection joins coalition to set standard for responsible government AI
Today, Reflection joined a coalition with AWS, Microsoft, OpenAI, SpaceX, Google, and Nvidia on a framework for how the U.S. government licenses and deploys AI – including a non-binding memorandum of understanding with the DoW.
What makes this notable for the community is what's written into the terms:
- Safety, red-teaming, and ongoing evaluation are committed practice, not aspirational language. We'll be part of that work.
- Explicit prohibitions on unlawful mass surveillance and autonomous weapon use.
- Constitutional and civil liberties protections in writing.
Responsible and lawful use isn't a footnote in this agreement, it's the foundation. That matters because it sets a precedent for how frontier AI gets safely deployed at scale, with evaluation and accountability as defaults rather than afterthoughts.
We continue to be committed to open source. Open releases, transparent safety work, and reproducible red-teaming remain the foundation of our plans. This agreement extends that posture into how our models can be used by the U.S. government including customizing them for scientists in national labs for their own research.