US and EU Sign First Transatlantic AI Safety Framework AISS: Covering ~800 Million People

According to combined reporting from Reuters and the Financial Times, the United States and the European Union have signed the first transatlantic AI safety framework, AISS. The framework covers approximately 800 million people and brings foundational AI systems and high-risk AI systems under pre-deployment safety review.

The most closely watched requirement of AISS is that covered systems must pass an independent third-party safety audit before deployment. Compared with self-assessment by companies, third-party audits mean model developers must provide external review bodies with sufficient materials to judge risk, and must meet a more unified safety threshold.

Which Systems Are Constrained

According to Reuters and the Financial Times, the framework focuses on "foundational" and "high-risk" AI systems. The former typically refers to underlying models that can be widely applied to different tasks, while the latter refers to applications that may have a significant impact on safety, rights, or important social activities.

• Pre-deployment audits: Rather than waiting for incidents to occur before assigning blame, checks are moved upstream.
• Independent assessment: Audit bodies must be independent from model development and deployment companies.
• Cross-regional coordination: The US and EU are attempting to establish a jointly enforceable safety baseline.

Market Effect of Covering ~800 Million People

The US and EU markets combined cover approximately 800 million people. For AI companies that want to serve customers in both regions, a transatlantic framework may reduce some duplicate compliance work; but pre-deployment audits will also add to testing, documentation preparation, and external assessment costs.

Analysis: The actual impact of AISS will depend on the specific ways "foundational" and "high-risk" are defined, the standards for third-party audits, and how US and EU regulators handle differences between existing regimes. The signing of the framework itself does not mean all implementation details have been clarified.

Chinese AI Companies Excluded

Comprehensive reporting notes that Chinese AI companies have been excluded from the framework's scope of participation. This makes AISS not only a technical safety arrangement but also one with clear geopolitical policy boundaries. For Chinese companies planning to enter US and EU markets, whether and how they can meet audit and market-access requirements in the future will require continued attention to subsequent implementation documents.

According to Reuters and the Financial Times, AISS requires foundational and high-risk AI systems to pass independent third-party safety audits before deployment.

Outlook: The most important points to watch next include how audit bodies gain recognition, what evidence companies need to submit, and how AISS connects with the EU's existing AI legal architecture.