A group of academics from University of the Sunshine Coast has published a study in the journal AI and Ethics, alerting businesses to the moral and technical risks of adopting generative AI.The study argues that generative AI, which can create realistic text, images, and audio, is seen as a “panacea of efficiency and effectiveness” by many organisations, but it also poses serious threats to privacy, security, and quality.
According to the study, some of the potential harms of using generative AI without proper safeguards include exposing sensitive data of the public, staff, and other stakeholders through mass data breaches or model extraction; making faulty decisions based on corrupted or manipulated AI models; and losing trust and reputation due to poor performance or embarrassing results.
“Generative AI is a powerful and promising technology, but it also requires careful and responsible use,” said Dr Declan Humphreys, co-author and lecturer in cyber security at UniSC.
“Many organisations are rushing to adopt generative AI without fully understanding its capabilities and limitations, or the ethical and cyber security implications. This can expose them to significant risks and liabilities”.
The authors suggest a five-point checklist for organisations that want to use generative AI in a secure and ethical way: they should design their AI models with security and ethics in mind; they should collect their data in a trusted and fair manner; they should store their data securely; they should follow ethical principles for retraining and maintaining their AI models; and they should train and manage their staff to use generative AI appropriately.
