S. 2177: No Adversarial AI Act

This bill, known as the No Adversarial AI Act, aims to protect the U.S. government from artificial intelligence (AI) systems developed by foreign adversaries. Here is a detailed breakdown of the bill's provisions:

Formulation of a Foreign Adversary AI List

The bill requires the Federal Acquisition Security Council to create a list of AI developed by identified foreign adversaries within 60 days of its enactment. This list must then be made public on a website within 180 days. The list will be updated at least every 180 days to reflect any changes or new developments. AI can be removed from the list if the owner certifies it was not developed by a foreign adversary and if this claim is reviewed and certified by the Federal Acquisition Security Council.

Prohibition on Use and Acquisition of Foreign Adversary AI

Executive agencies are instructed to review AI listed as being developed by foreign adversaries and to consider barring its use within 90 days of the bill's enactment. They must employ certain authorities to ensure that AI from these adversaries is excluded or removed from use. However, there are exceptions that can be approved after notifying specific congressional committees if the use of such AI is deemed necessary for:

• Scientifically valid research
• Evaluation, training, testing, or analysis
• Counterterrorism or counterintelligence activities
• Ensuring critical mission functions are not jeopardized

Definitions and Terms

The bill provides specific definitions for terms, including:

• Artificial Intelligence: As defined in prior U.S. legislation, encompassing systems and techniques related to AI.
• Executive Agency: Any agency as defined under U.S. law.
• Foreign Adversary: A term defined in U.S. law that identifies specific countries deemed adversarial.
• Foreign Adversary Entity: This includes individuals or entities associated with foreign adversaries, such as companies incorporated in these countries or those with significant foreign ownership.

Implementation Timeline

The bill lays out a timeline for its implementation, including:

• 60 days for the creation of the foreign adversary AI list
• 180 days for the publication of the list
• 90 days for executive agencies to begin reviewing AI systems for potential bans

Oversight and Accountability

The bill requires ongoing oversight of the AI systems in question, ensuring that only those not associated with foreign adversaries are used by U.S. government agencies. This is intended to mitigate potential security risks that could arise from using foreign-developed AI technologies.

Relevant Companies

• MSFT - Microsoft: As a leader in AI development, any AI technology developed by Microsoft that could potentially involve foreign partnerships may need scrutiny under this bill.
• GOOGL - Alphabet Inc.: Similar to Microsoft, Google's AI projects might be impacted if any of their AI systems are deemed developed by foreign adversaries.
• AMZN - Amazon: Any AI technology developed by Amazon for its services, particularly in cloud computing, could be affected if tied to foreign entities.