State attorneys general continued using existing consumer-protection, privacy, discrimination and deceptive-practices laws to oversee AI, even without comprehensive state AI statutes. Reuters
The growing state-by-state regulatory patchwork remained a major compliance issue, particularly for automated employment decisions, healthcare, insurance, deepfakes and high-risk consumer systems.
Federal proposals that could preempt portions of state AI law continued generating debate over whether AI should be governed nationally or through state-level protections.
US National Policy
Members of Congress prepared the bipartisan AI Kill Switch Act, which would authorize emergency throttling or shutdown of advanced systems in severe loss-of-control situations. The Verge
Lawmakers also discussed independent security evaluations, mandatory incident reporting and shutdown controls for frontier models following the OpenAI agent-containment failure.
The United States and China reportedly planned government-level AI talks for September covering frontier models, cybersecurity, military risk and strategic competition. Reuters
EU AI Policy
The European Commission published guidance on AI Act transparency obligations for providers and deployers of certain AI systems on July 20.
The guidance supports approaching August compliance milestones affecting generated-content disclosure, user notification and transparency around certain AI interactions.
The EU also advanced its Cybersecurity and AI Action Plan, including expanded European capacity to evaluate advanced models before deployment. European Commission
Chinese AI policy
China reportedly considered tighter export controls on advanced AI models and semiconductor technologies, treating both as strategic national assets. Reuters
Beijing continued promoting open-source AI as a way to increase access for developing countries and strengthen China’s influence over international AI standards.
Planned U.S.–China discussions demonstrated that frontier-model safety, military applications and cyber capabilities are becoming formal diplomatic issues.
NIST
No major new NIST AI publication was identified during the July 20–26 window.
NIST’s AI Risk Management Framework remained the primary federal reference for mapping, measuring, managing and governing AI risk. NIST AI RMF
The OpenAI containment incident reinforced the need to extend risk-management practices to autonomous actions, cyber capabilities, connected tools and environmental controls.
US Government
The proposed AI Kill Switch Act would give the Department of Homeland Security emergency authority over advanced systems under narrowly defined catastrophic conditions.
The administration expanded a voluntary pledge intended to prevent electricity costs associated with AI data centers from being shifted to residential consumers.
Federal departments received direction to align future research investments around AI, advanced computing, energy, national security and foundational science.
State Government
Governors and state utility organizations increasingly addressed the electricity, water, land-use and cost-allocation effects of AI data-center expansion.
States continued evaluating protections that would prevent households from bearing infrastructure-upgrade costs created by large data centers.
State governments also expanded AI inventories, workforce policies, procurement controls and public-sector governance programs.
Law Enforcement
Facial-recognition deployment remained the leading law-enforcement AI issue. Reporting from the United Kingdom showed increasing arrests resulting from live facial-recognition scans, accompanied by concerns about proportionality, bias and the lack of comprehensive legislation.
Western Australia began a facial-recognition trial that scanned more than 130,000 people and produced 19 arrests during its initial period, alongside two false alerts.
These developments reinforced the need for human verification, warrant and retention policies, accuracy testing, appeal mechanisms and public disclosure.
Global Regulation and Global Government
U.S.–China AI talks planned for September demonstrated that frontier-model governance is becoming part of formal international diplomacy.
China promoted a more open and internationally accessible AI-development model while simultaneously considering restrictions on exporting its most advanced models and chips.
Technology companies pushed back against broad restrictions on open-source AI, arguing that open models support competition, sovereignty and lower deployment costs. Reuters
Governments continued debating whether advanced AI should be governed through voluntary standards, licensing, pre-deployment evaluations, incident-reporting obligations or mandatory shutdown mechanisms.
