Secure AI development guidance: What software teams need to know

The report outlines new NCSC and CISA guidelines for secure AI system development, endorsed by 16+ nations and major industry organizations, emphasizing a Secure by Design approach across four pillars: design, development, deployment, and operations. It stresses transparency and accountability via comprehensive documentation (model/data cards, prompts, SBOMs), addresses adversarial ML risks and prompt injection, and argues that AI providers should bear the burden of security defaults and disclosure across the supply chain. Industry leaders welcome the guidance while calling for stronger focus on model interpretability to better manage AI risks and support secure innovation.