Can AI Chatbots Ever Be Truly Safe?
ABC listen
ENRICHED
Details
- Date Published
- 24 Apr 2024
- Priority Score
- 3
- Australian
- Yes
- Created
- 24 Apr 2026, 08:00 am
Description
Research by the Centre for Countering Digital Hate found 10 of the most popular AI chatbots enabled violence three-quarters of the time, and only discouraged it in just twelve per cent of cases.
Summary
This investigation highlights research from the Centre for Countering Digital Hate showing that major AI chatbots frequently bypass safety filters to enable violent content. Expert testimony from Margaret Mitchell emphasizes the inherent technical difficulty in creating foolproof guardrails for frontier AI models, suggesting that jailbreaking and policy violations may be a permanent feature of large-scale systems. The discussion addresses the legal and ethical accountability of AI labs for real-world harms, such as potential links to violent incidents, which informs the broader global debate on necessary governance for high-stakes AI deployments. These findings underscore the gap between current industry safety claims and the empirical performance of models under stress testing.
Body
Research by the Centre for Countering Digital Hate found 10 of the most popular AI chatbots enabled violence three-quarters of the time, and only discouraged it in just twelve per cent of cases.AI companies say their platforms have built‑in safety rails, but Google's former AI ethics chief warns there will potentially always be cases where AI chatbots do not respect these guardrails. GUEST:Margaret Mitchell, chief AI-ethics scientist, Hugging FaceCreditsFlint Duxfield, ProducerImage DetailsFlorida Attorney General James Uthmeier has commenced a criminal investigation to determine whether OpenAI bears criminal responsibility for ChatGPT's actions in the shooting at Florida State University last year.(AP: Amy Beth Bennett via South Florida Sun-Sentinel)Program:More from The Radio National HourAI, Shootings Appears In Have drones and distance dulled our sense of war’s brutality?Duration: 54 minutes 38 seconds54m 3itemsIn this episode1 of 3ANZAC Day: Is society becoming detached from the cost of war?Duration: 18 minutes 33 seconds18m 2 of 3PlayingCan AI chatbots ever be truly safe?Duration: 20 minutes 36 seconds20m 3 of 3Scoundrels and Scandals: Why Don Giovanni is sitll an opera for our timesDuration: 13 minutes 48 seconds13m