Member

The Online Safety Act currently focuses on content detection and moderation. Adding AI chatbots to the Act therefore requires a different approach, as these systems have the ability to generate such content in the first place. The amendment from the noble Baroness, Lady Kidron—hallelujah for her—would ensure that this risk assessment process tackles the distinct risk of AI platforms. This should involve auditing training data and testing systems, using methods such as red-teaming, to ensure that models cannot be exploited to create child sexual abuse material. Only after we know that these models are safe should they be released to the public.