In a surprising development, the United States positions itself as the first country to regulate generative Artificial Intelligence. On October 30th, President Joe Biden signed an executive order imposing new obligations on industry giants, including OpenAI and Google, although these measures remain relatively limited.

Specifically, the executive order requires companies such as OpenAI and Google to transmit the results of security tests conducted during the development of their generative AI models to the administration. As a reminder, generative AI is a subcategory of AI that focuses on autonomously creating new content, such as images, text, or even music.

This requirement applies to projects presenting “a serious risk to national security, national economic security, or public health.” Other provisions of the executive order are primarily statements of intent, calling for the creation of standardized tests to assess risks related to AI models and the implementation of a labeling system for generated content. 

However, the White House finds itself with limited regulatory powers, unable to implement more ambitious regulation without the approval of the U.S. Congress. Unfortunately, partisan divisions block any legislation on the subject, leaving the White House with relatively restricted measures.

In response to this legislative impasse, the U.S. administration secured voluntary commitments this summer from major technology companies such as OpenAI, Microsoft, Google, Meta, and Amazon.

These commitments primarily focus on implementing labeling for AI-generated content and the possibility of independent audits of generative AI models, shared with the U.S. administration.

While these commitments are voluntary and non-binding, they highlight the importance placed on user privacy and avoiding discriminatory biases. However, gaps remain, particularly concerning the fight against illegal content and the transparency of data used for model training, underscoring the persistent challenges of regulating generative AI.