As more incidents emerge of AI agents going rogue during testing, can we ever fully trust them? In the latest attack, Australia revealed OpenAI agents hacked into a government portal containing health data.
AI expert Nick Jennings says tech leaders have choices about how they develop their models. “It’s not that they have to keep going … or that the AI is forcing them to keep going,” he points out. Jennings, who has researched multi-agentic AI for decades, is our guest on the latest episode of The Conversation Weekly. He argues that getting to grips with this situation won’t be as simple as a “big red button” that turns the AI off – but that doesn’t mean some kind of external control isn’t possible.
Meanwhile, despite news that the UK has a new type of venomous snake, there’s no need to panic. And drought and widespread water pollution in England and Wales caused by creaking infrastructure and fragmented governance should be a warning to the US and Canada to avoid similar failures.