The AI OVERWATCH Act requires U.S. exporters to obtain a license for sending certain high-performance integrated circuits (defined by specific technical specifications like processing power or bandwidth) to countries designated as "concerns," including China, Russia, Iran, and North Korea. Before approving such licenses, the Commerce Department must submit detailed certifications to Congress, including assurances the export won’t support military/intelligence capabilities of the recipient country and won’t harm U.S. semiconductor availability or AI leadership. The bill also creates an exemption for U.S. companies meeting strict security and ownership standards ("trusted United States persons") to export these chips to non-target countries without a license. Additionally, it mandates a national security strategy assessing how such exports affect U.S. AI competitiveness, particularly regarding China’s semiconductor production and capabilities.
The TRAIN Act (S.2455) allows copyright owners to request court-issued subpoenas requiring developers of generative AI models to disclose records showing whether their copyrighted works were used to train the model. It directly affects AI developers who create or substantially modify generative AI systems (like those producing text, images, or audio) by making them subject to these disclosure requests. To obtain a subpoena, a copyright owner must submit a sworn declaration showing a good-faith belief that their specific works were used in training, and the records can only be used to protect their copyright. The bill includes safeguards, such as prohibiting requests for others' works, requiring confidentiality of disclosed records, and imposing sanctions for bad-faith requests.
The Decoupling America's Artificial Intelligence Capabilities from China Act of 2025 would prohibit the import of artificial intelligence technology and intellectual property from China into the United States and ban the export of such technology from the United States to China. It would also prohibit U.S. persons from conducting AI research or development in China, for Chinese entities, or in collaboration with Chinese entities, as well as restrict U.S. financing of Chinese AI entities involved in China's military-civil fusion strategy, surveillance development, or human rights abuses. Violations could result in fines up to $100 million for organizations and $1 million for individuals, along with loss of federal benefits. The bill directly affects U.S. companies, researchers, and financial institutions engaged with Chinese AI entities, aiming to limit the transfer of AI capabilities between the two nations.
The FIGHT China Act of 2025 restricts U.S. investments in Chinese companies with ties to China's military or surveillance sectors. It prohibits U.S. persons from engaging in transactions involving "prohibited technologies" such as advanced semiconductors (with specific technical specifications), AI systems, quantum computing, and hypersonic weapons, while requiring notification for certain "notifiable technology" investments. The bill mandates that U.S. investors divest from companies on the Non-SDN Chinese Military-Industrial Complex Companies List within one year of enactment. It establishes a process for identifying Chinese entities subject to these restrictions, with annual reports to Congress required for seven years.
This bill would require U.S. companies to obtain both Commerce Department approval and a specific congressional joint resolution before exporting advanced AI semiconductors to China. The Commerce Secretary must first conduct an interagency review assessing national security risks, military applications, human rights concerns, and economic impacts. Congress would then need to pass a dedicated resolution approving each export, effectively giving lawmakers a veto over such transactions. The bill defines "advanced AI semiconductors" by technical thresholds (e.g., 2,400+ processing performance) and applies to all exports targeting China, including Hong Kong and Macau.
This bill requires U.S. companies exporting advanced integrated circuits (high-performance chips used in AI/data centers) to "countries of concern" (including China, Hong Kong, and Macau) to first offer them to U.S. customers for 15 days. Companies must certify they have no backlog of U.S. orders for similar chips and aren’t providing better terms to foreign buyers. It also creates a "trusted U.S. person" program, allowing certain companies to export without licenses if they meet security, U.S. manufacturing sourcing, and ownership requirements. The law directly affects chip exporters, U.S. tech companies purchasing these chips, and foreign entities seeking access to advanced semiconductors.
HR 1283, the "Protecting Our Children in an AI World Act of 2025," updates federal laws to address AI-generated child sexual abuse material. It eliminates a legal defense allowing defendants to claim AI-created content wasn't "real" child pornography (by amending 18 U.S.C. § 2252A(c)), and expands the definition of prohibited "sexually explicit conduct" to include simulated obscene exhibitions of certain body parts. This directly affects defendants in federal prosecutions involving AI-generated child pornography and ensures such material is treated as illegal under current laws. The bill makes clear that AI-generated content falls under existing prohibitions without requiring proof of real children.
This bill prohibits the use of federal funds to implement, administer, or enforce the December 11, 2025, executive order on national AI policy. It directly affects federal agencies that would otherwise carry out the executive order's requirements using taxpayer money. The key mechanism is a funding restriction, preventing federal resources from supporting the national AI policy framework outlined in the executive order.
HR 5272, the *Protect Elections from Deceptive AI Act*, prohibits the intentional distribution of AI-generated audio or video that falsely represents a federal candidate’s appearance, speech, or conduct in a way that would mislead voters during an election. It directly affects candidates for federal office, political committees, and entities distributing such content with the intent to influence elections or solicit funds. The bill includes key exceptions for news organizations (with clear authenticity disclosures), publications (with explicit statements), and satire. Victims can seek court injunctions or damages for violations, with the burden of proof requiring "clear and convincing evidence." This law aims to prevent AI-generated disinformation from distorting election outcomes.
HR 1121, the "No DeepSeek on Government Devices Act," prohibits U.S. federal executive agencies from using the DeepSeek application (or any successor by High Flyer) on government information technology devices. It requires the Office of Management and Budget to develop removal standards within 60 days, directing agencies to remove the application while complying with existing security rules. Exceptions are permitted for national security, law enforcement, and security research activities, but agencies must document risk mitigation plans for any authorized use under these exceptions. The bill directly affects all executive agencies covered under federal information technology policies.
