Introduction#
The White House is exploring the possibility of an executive order that would create a system for testing new artificial intelligence (AI) models before they are made available to the public. This initiative aims to ensure that these technologies are safe and do not pose risks to businesses or government operations.
Proposed Testing Framework#
According to Kevin Hassett, the Director of the National Economic Council, the proposed framework would require AI models to undergo rigorous testing similar to the approval process used by the Food and Drug Administration (FDA) for new drugs. This means that before any AI model is released, it would need to be thoroughly evaluated to confirm its safety and effectiveness.
Recent Concerns and Discussions#
The move comes in response to concerns raised by Anthropic PBC regarding its AI model, Mythos, which has the capability to identify vulnerabilities in networks. This has raised alarms about potential cybersecurity risks. As a result, access to Mythos has been limited to select large tech and financial firms. The Trump administration had previously sought to test this model within federal agencies, a plan that Hassett has confirmed is still in discussion.
Broader Implications#
Hassett noted that if the executive order is implemented, the testing requirements would likely apply to all AI companies. This marks a significant shift from the previous administration's approach, which favored minimal regulation of AI technologies. Additionally, the Commerce Department has recently expanded a voluntary AI testing program, with major companies like Google and Microsoft agreeing to provide access to their models for government evaluation and improvement of security measures.