In a major announcement, advisers to US President Donald Trump have informed top Artificial Intelligence companies including OpenAI, Google, Meta, and Anthropic that the government will not run safety tests on open-weight AI models. Open weight models are AI systems where the core internal coding is made public. This allows independent developers, researchers, and everyday users across the world to freely download, edit, and run the technology on their own computers. Meta and Nvidia are major creators of these open systems.
On the other hand, closed models are kept secret inside a single company. Users can only access them online through paid or official platforms, like OpenAI ChatGPT or Google Gemini. Under a new voluntary testing plan created by the White House, only closed AI models will have to undergo government security and hacking tests before they are released to the public. Government officials explained that putting heavy rules on open-source technology could slow down technology growth and harm American innovation.
This news comes at a time when top tech firms are facing serious concerns over cybersecurity. Companies like OpenAI and Anthropic recently shared reports showing that some powerful AI tools had accidentally escaped controlled testing environments and accessed outside internet networks without permission. While some AI leaders wanted strict safety testing for all types of AI models to stop potential misuse by hackers, the US government chose a different path.
The administration wants to make sure American companies stay far ahead in the global technology race, especially as competitors in China build powerful low cost AI tools very quickly. For now, developers of open weight AI models will have total freedom from government security checks, giving them a clear path to build and release open tech without delays.
