NCSC Guidelines on Secure AI
The National Cyber Security Centre has published guidelines for secure AI development.
Below are the key takeaways:
Secure design: (1)raise staff awareness of threats and risks, (2)model the threats to your system, (3)design your system for security as well as functionality and performance, and (4)consider security benefits and trade-offs when selecting your AI model.
Secure development: (1)secure your supply chain, (2)identify , track and protect your assets, (3)document your data, models and prompts, and (4)manage your technical debt.
Secure deployment: (1)secure your infrastructure, (2)protect your model continuously, (3)develop incident management procedures, and (4)release AI responsibly and make it easy for users to do the right things.
Secure operation and maintenance: (1)monitor your system's behaviour, (2)monitor your system's input, (3)follow a secure by design approach to updates, and (4)collect and share lessons learned.
The full guidelines can be found here.