Wednesday, June 25, 2025

The No Adversarial AI Act 2025

The 2025 AI Index Report by Stanford University’s Human-Centered Artificial Intelligence center has the US in the lead in producing top AI models. 
But the report notes China is rapidly closing the performance gap, reaching near parity in 2024 on several major benchmarks. It also shows that China leads in AI publications and patents.
US Lawmakers Introduce 'No Adversarial AI Act' To Blacklist Chinese AI From  Federal Agencies
AI models need more standards and tests, say researchers
AI models need more standards and tests, say researchers

US lawmakers introduce bill to bar Chinese AI in US government agencies

FILE PHOTO: Illustration shows message reading "AI artificial intelligence" and robot hands · Reuters

SAN FRANCISCO (Reuters) - A bipartisan group of U.S. lawmakers on Wednesday planned to introduce a bill in both houses of Congress that would bar U.S. executive agencies from using artificial intelligence models developed in China, including those from DeepSeek.

The introduction of the bill, dubbed the "No Adversarial AI Act," comes after Reuters reported that a senior U.S. official has concluded that DeepSeek is aiding China's military and intelligence operations and has had access to "large volumes" of Nvidia's chips.

DeepSeek shook the technology world in January with claims that it had developed an AI model that rivaled those from U.S. firms such as ChatGPT creator OpenAI at much lower cost. Since then, some U.S. companies and government agencies have banned the use of DeepSeek over data security concerns, and President Donald Trump's administration has mulled banning its use on U.S. government devices.

  • The bill introduced Wednesday into the U.S. House of Representatives by Representative John Moolenaar, a Michigan Republican who chairs the Select Committee on the Chinese Communist Party, and Representative Raja Krishnamoorthi, an Illinois Democrat who is the ranking member on the committee, would create a permanent framework for barring the use of all Chinese AI models from U.S. executive agencies, as well as those from Russia, Iran and North Korea.
  • The bill would require the Federal Acquisition Security Council to create a list of AI models developed in those countries and regularly update it.

    1. Federal agencies would not be able to buy or use those AI technologies without an exemption, such as for carrying out research, from the U.S. Congress or the Office of Management and Budget. 
    2. The law also contains a provision that can be used to get technologies off the list with proof that they are not controlled or influenced by a foreign adversary of the U.S.

57 minutes ago  What the No Adversarial AI Act Does: Creates a public list of AI systems developed by foreign adversaries, maintained and updated by the ...
7 minutes ago  The No Adversarial AI Act, as proposed, seeks to identify AI systems developed by foreign adversaries and ban their use in the U.S. government, ...
1 hour ago  Federal agencies would be barred from using artificial intelligence linked to the Chinese government under bipartisan legislation introduced ...
2 hours ago  The bill aims to block not just Chinese AI tools but also those from Russia, Iran and North Korea from being used in the U.S. government.
3 hours ago  A bipartisan group of lawmakers is introducing new legislation to block federal agencies from buying or using artificial intelligence ...
Select Committee on the Chinese Communist Party (@committeeonccp) / X 
Important Notes:
  • This is a proposed piece of legislation and its passage is not guaranteed.
  • The bill includes exceptions for research, counterterrorism, or mission-critical needs, with potential exemptions granted by Congress or the Office of Management and Budget.
  • The Senate version of the bill is spearheaded by Sens. Rick Scott and Gary Peters. 
 

https://blog.talosintelligence.com/content/images/2025/06/LLM-abuse.jpg

Cybercriminal abuse of large language models


No comments: