Geely Auto Group: Building Certified AI Safety into Vehicle Development

Geely Auto Group has become the first global automaker to achieve ISO/PAS 8800 certification for its AI safety engineering processes, following independent evaluation by SGS-TÜV. The certification confirms that Geely's defined development capability – spanning functional safety, expected functional safety and AI safety – meets the requirements of three converging international standards, as intelligence becomes increasingly central to vehicle value.

About Geely Auto Group

Geely Auto Group is a leading automobile manufacturer headquartered in Hangzhou, China. Founded in 1997 as a subsidiary of Zhejiang Geely Holding Group, the Group manages several major brands, including Geely Auto, Lynk & Co and Zeekr, and is the global strategic partner of Malaysian national automaker PROTON.

A new standard, and the challenge of catching up fast

As artificial intelligence becomes increasingly central to vehicle value and brand differentiation, Geely set out to build a systematic capability for the safe application of AI in its vehicles, aligned with the newly published ISO/PAS 8800.

Geely had already established functional safety processes under ISO 26262 and SOTIF processes under ISO 21448 at concept and system level, with these processes certified by UL. Building on this foundation, Geely moved into the software phase and expanded into AI safety in line with ISO 8800:2024, developing new processes and seeking third-party certification to catch up quickly with the latest standard.

Closing the gap: how we supported the certification journey

We carried out a gap analysis of Geely's development processes against the requirements of ISO/PAS 8800, delivered by two experts at SGS-TÜV Saar in Germany. The assessment covered:

• Identification and evaluation of the current status against functional safety requirements, including strengths and weaknesses
• Determination of the activities needed to achieve compliant, safety-related development, with prioritization
• Documentation of the evaluation results in a detailed protocol

The gap analysis did not include inspection services for protection against product liability claims; these form part of separate follow-up evaluation and certification services. The consulting activities were carried out by our consulting team in Shanghai.

The inspection itself was based on the requirements of the normative reference, independently evaluated. It included a safety audit, introducing and inspecting the process environment under consideration, followed by inspection of Geely’s process descriptions. These needed to comply with general standards as well as configuration and change management rules.

Results were documented in protocols, giving Geely the opportunity to correct any findings identified. Corrections were checked by SGS-TÜV Saar, with further reviews falling outside the scope of this project. Before release of the evaluation report the complete documentation had to be available, in line with the relevant standards.

What certification means for Geely

The engagement concluded with an evaluation report produced by our inspectors, covering the defined development processes and engineering scope assessed. Building on this, Geely is now positioned to carry out a further product audit and assessment, which can in turn support a project-dependent certificate for specific vehicle programs.

Setting the benchmark in automotive AI safety

Announced in Hangzhou on June 24, 2025, Geely Auto's ISO/PAS 8800 certification marks it as the first global automaker to certify its AI safety engineering processes against this standard. The certification applies to the defined development and engineering scope assessed, rather than to every vehicle or AI system in Geely's current lineup.

The accreditation, from SGS-TÜV Saar in Germany, confirms Geely's engineering expertise across three key international standards: functional safety, expected functional safety and AI safety.

ISO/PAS 8800, introduced by the International Organization for Standardization in December 2024, is the first globally authoritative standard for artificial intelligence safety in the road vehicle sector – marking a shift in automotive AI safety from point-in-time safety compliance toward a comprehensive, full-lifecycle approach.