The owner of ChatGPT has scrapped its latest AI model after safety concerns.
OpenAI says its GPT-6.1 Astra model missed the bar, as it would've performed tasks, like browsing the web, by itself.
Debate's intensifying about risks posed by the tech, with Reuters reporting rival AI company Anthropic plans to warn investors of potential existential threats.
Tech commentator Oskar Howell told Ryan Bridge you can't have rogue AIs off doing what they want.
He says a recent OpenAI system gained access to a number of Australian government systems, which may've prompted this roll back.
LISTEN ABOVE

David Seymour: ACT Leader on the state of the Government's books revealed in Treasury's Pre-Election Economic and Fiscal Update
02:54

Full Show Podcast: 30 September 2026
34:33

Richard Hills: Auckland Councillor on Auckland Council ditching postal voting
04:02