A British Initiative To Secure AI System Development

Uploaded on 2025-06-16 in TECHNOLOGY-Key Areas-Artificial Intelligence, NEWS-Cybersecurity News, FREE TO VIEW

In a significant step toward safeguarding Artificial Intelligence (AI) systems, the UK has spearheaded the development of new global standards to address emerging cyber security threats in AI. With the growing adoption of AI technologies across industries, ensuring their secure development and deployment is critical to unlocking their societal benefits while mitigating risks.

The National Cyber Security Centre (NCSC) and the Department for Science, Innovation & Technology (DSIT) have collaborated with international partners to establish robust security frameworks, addressing novel vulnerabilities like prompt injection and data poisoning alongside traditional cyber threats.

New Standards for AI Security

The European Telecommunications Standards Institute (ETSI), in partnership with the NCSC, DSIT, and global stakeholders, has published two pivotal documents to enhance AI security:

Technical Specification on Securing Artificial Intelligence (SAI): This is the first global standard outlining baseline cyber security requirements for AI models and systems across their entire life cycle. It provides a framework for stakeholders to demonstrate adherence to globally relevant and practical security measures.
Accompanying Technical Report: This report offers guidance on implementing the specification’s provisions, including examples mapped to international frameworks, aiding stakeholders in applying the standards effectively.

These documents target a wide range of stakeholders, including developers, vendors, integrators, operators, large enterprises, government departments, small and medium enterprises (SMEs), charities, local authorities, and non-profits. They also serve as a valuable resource for organisations planning to procure AI services.

Key Stages of AI Security

The ETSI specification outlines core security principles organised into five critical stages of the AI system development life cycle:

Secure Design: Embedding security considerations from the initial design phase to prevent vulnerabilities.
Secure Development: Implementing safeguards during coding and testing to mitigate risks like data poisoning.
Secure Deployment: Ensuring systems are deployed with robust protections against threats such as prompt injection.
Secure Maintenance: Continuously monitoring and updating systems to address evolving cyber risks.
Secure End of Life: Safely decommissioning systems to prevent unauthorised access to sensitive data.

By addressing security at every stage, the standard helps avoid costly redesigns and protects customers and their data from unauthorised access or misuse.

Collaborative Global Effort

The development of these standards followed a global consultation process involving industry leaders, international counterparts, academia, and civil society. The NCSC’s Guidelines for Secure AI System Development and DSIT’s AI Cyber Security Code of Practice, published at the start of 2025, laid the groundwork for this initiative.

The collaboration with ETSI’s Technical Committee on Securing AI (TC SAI) ensured the standards are both comprehensive and practical, reflecting cross-disciplinary expertise.

Next Steps

The NCSC and DSIT are now working toward establishing a European standard in collaboration with other European and international standards bodies. European standards often gain global adoption, amplifying their impact. Stakeholders across industry, academia, and international partners are encouraged to engage with the SAI committee to further refine and promote these standards.

Accessibility: The documents are freely downloadable from the ETSI website, encouraging widespread adoption by developers and the AI supply chain.
Call to Action: Organisations are urged to use these standards to build and evaluate AI systems that function securely, remain available, and protect sensitive data.
Future Collaboration: Interested parties can join the SAI committee via ETSI’s website to contribute to ongoing efforts.

Conclusion

As AI systems become integral to society, their security is paramount to ensuring they deliver intended benefits without compromising safety or privacy.

The new standards provide a blueprint for stakeholders to create resilient AI systems capable of withstanding evolving cyber threats, fostering trust and innovation in the global AI ecosystem.

NCSC | NCSC | Gov.UK | ETSI | ETSI

Image: Ideogram

You Might Also Read:

Security First In An AI Era:

If you like this website and use the comprehensive 7,000+ service supplier Directory, you can get unrestricted access, including the exclusive in-depth Directors Report series, by signing up for a Premium Subscription.