US, Australian, New Zealand and UK cybersecurity agencies publish guidance on best practices for securing data used to train and operate AI systems

Summary

On May 22, 2025, cybersecurity agencies from the US, UK, Australia, and New Zealand released a Cybersecurity Information Sheet (CIS) focused on securing data for AI systems. This guidance highlights the significance of protecting data integrity to ensure reliable AI processes. It identifies critical risks such as data supply chain vulnerabilities, maliciously altered data, and data drift, while providing strategies for organisations to mitigate these risks when deploying AI technologies.

Source: Lexology

Key Points

• Cybersecurity agencies advocate for best practices in data security for AI systems.
• Risks include inaccuracies from third-party data, “poisoned” data, and data drift.
• Recommendations include using trusted data sources, maintaining data integrity, and implementing regular data checks.
• Data encryption and secure data handling practices are emphasised.
• Regular risk assessments using established frameworks are crucial for organisational security.

Why should I read this?

If you’re involved in AI development or data management, this article is a must-read! It’s packed with essential guidance from top cybersecurity agencies, helping you navigate the crucial landscape of data security for AI systems. Trust us, a few minutes here can save you a headache down the line – don’t skip it!