MC21 INCYBER Ensuring Security to Enable AI Safety

In the landscape of ultra rapid evolving technologies due to the power of generative AIs, all AI systems are increasingly vulnerable to diverse cybersecurity threats, ranging from data breaches, data poisoning, to adversarial attacks. Establishing robust security measures not only safeguards AI systems against malicious intrusions but also bolsters overall trust in the systems’ reliability and efficacies. Within this framework, data integrity emerges as the cornerstone component. Maintaining data integrity ensures that the information AI systems rely upon remains accurate, consistent, and free from tampering and poisoning throughout the lifecycle, which is a big challenge in any organization’s operational environment. By fortifying the integrity of data through steganography watermarking of all data types, organizations can bolster the resilience of AI systems against manipulation and exploitation thereby advancing the broader goal of ensuring AI safety. This talk underscores the critical interplay between AI security and data integrity, illuminating their indispensable roles in fostering trust and resilience in an increasing AI-driving world.