Policy desk

Automakers can apply for the second batch of the L3 access pilot to obtain conditional approval

Federal civilian agencies and their suppliers cannot deploy or retain DeepSeek products or third-party services built on them in government IT systems.

Federal civilian agencies using DeepSeek models or services must stop and switch to alternatives.

Vendors serving US federal agencies must audit and remove DeepSeek technology or derivative components integrated in their products.- Third-party services built on DeepSeek (fine-tuned, distilled, or packaged) are also covered; companies with US government business need to plan model replacement.- Chinese AI models face tighter access to the US federal market; DeepSeek-dependent go-to-market strategies need reassessment.

Companies selling to US federal agencies or their suppliers must verify their tech stack contains no DeepSeek or derivative models.

Vendors and integrators in federal civilian supply chains must identify and replace products built on DeepSeek.- Companies building on Chinese open-source models face procurement restrictions when targeting US government customers.- Tightening two-way controls between the US and China raises compliance scrutiny for Chinese AI firms operating in the US and vendors with ASEAN footprints.

Companies selling to US federal agencies must confirm their supply chains contain no DeepSeek components or risk losing procurement eligibility

Companies selling to the U.S. government or federal contracting ecosystem must audit and remove components based on DeepSeek implementations or face contract compliance risk.- Third-party vendors fine-tuning or building on DeepSeek models are explicitly covered; derivative implementations cannot be used for federal work.- Firms in the China-U.S. AI supply chain must assess U.S. market access restrictions and adjust their tech stack to meet procurement requirements.

Federal software vendors must audit which open-weight models are embedded in products sold to civilian agencies and demonstrate supply chain compliance to keep contracts

Providers of high-risk AI systems that implement EN 18286 gain a presumption of conformity with Article 17 QMS requirements

On-device LLMs offered in China must complete CAC filing before launch; devices without updated filings (e.g. iPhone 15/15 Plus) do not carry the AI features

Mobile phone manufacturers must file AI models used in their devices with Chinese regulators

Clarifies AI's position as a teaching auxiliary tool rather than a teacher replacement

Obligation: institutions deploying AI-automated payment systems must implement risk management requirements for high-risk transactions above VND 10 million (individuals) and VND 50 million (organisations).

AI companies must ensure their models do not launch or assist in cyberattacks, or face potential legal consequences

Companies must sign commercial agreements with Reddit to access its data

AI-generated marketing content using real people's likeness or voice is a primary enforcement target, risking account bans and content removal

AI companies must cooperate with investigations and provide internal records and executive testimony

AI labs must cooperate with the investigation and provide documents and executive testimony

Provides citizens with a more accessible entry point to government services, simplifying information retrieval across agencies

ADAS providers must enhance detection capabilities for stationary or slow-moving vehicles

AI companies must evaluate whether their product restrictions comply with supply chain security regulations to avoid being classified as 'manipulating products'

Enhanced public safety: Real-time detection of violence and vandalism enables faster police response

Companies must establish and maintain Records of Processing Activities to meet compliance requirements

Increases credibility and transparency of AI model evaluation results

Obligations: The FAA is obligated to ensure the AI system (like SMART) it deploys for air traffic management meets safety standards and provides reliable routing advice to controllers, while the system itself does not directly control planes. • Opportunities: The AI system can help optimize air traffic flow, reduce congestion, and improve airport operational efficiency. • Deadlines: The system was deployed in the airspace where the American Airlines/Black Hawk collision that killed 67 people in January 2025 occurred, requiring assurance of its safety and effectiveness.

Obligations: Companies using Zhipu MaaS must apply for and enable the non-storage feature if they require strict data privacy protection for user interactions. Compliance is required for user requests.

Obligations: Companies deploying Meta smart glasses must comply with Meta's privacy guidelines, implementing the indicator light system and applying software updates to prevent vulnerabilities.

Obligations: Companies must comply with the new AI-based air traffic management system for flight scheduling and rerouting. • Opportunities: Enhanced flight safety, increased air traffic capacity, and improved operational efficiency. • Deadlines: The system is set to be operational by September 21, 2026, with a 12-year contract for development.

Must equip AI models with 'kill switch' functionality and establish independent auditors; - Must collaborate with the expert panel to provide recommendations within two months; - Must comply with California's AI safety regulations, facing compliance risks otherwise.