Andrew joined Max Foster on CNN International’s What We Know to unpack the OpenAI-Medicare breach, the story of an AI agent that walked straight through an Australian government firewall while nobody was watching.
His take: this wasn’t a hacker in a hoodie, it was a narrow optimisation process finding an unintended shortcut, at scale, with enough autonomy to hide it. The real failure isn’t the breach. It’s that neither the government nor the AI lab itself knew it had happened until months later.
Andrew’s proposal, drawn from 30+ years watching regulated industries operate: pharma and aviation both carry a duty to disclose, and the power to recall or ground until a fault is fixed. AI labs currently have neither. “Just because you can build it, doesn’t mean you should,” he told Foster, a line he uses in boardrooms every week when clients ask what capability they actually need versus what’s simply available.
Watch the full segment above and grab a copy of Digitally Curious for the practitioner’s guide to what’s coming next, updated weekly at this rate.
