Is anybody really in charge? Why Angus Taylor might be ready to seize his last chance

September 28, 2026 — 3:00am

You have reached your maximum number of saved items.

Remove items from your saved list to add more.

It began to feel, last week, as though nobody were in charge – of one of the planet’s largest AI companies, of government agencies and of the world into which we seem to be heading.

In June, an autonomous artificial intelligence “agent” from OpenAI made it into supposedly private Medicare systems. In August, OpenAI finally discovered this breach. In September, it told Australia.

How did this world-conquering company choose to do this? By sending an email to a public email inbox at Services Australia. Another striking fact: until then, the government had been unaware its systems had been infiltrated.

There’s worse. According to Reuters, the company still isn’t sure it knows everything it should: it’s checking. What it does know is not good. On Friday, it revealed its agents had made it not just into Medicare but into websites of the American government. They had also leaked more than 50 images from users. On Sunday morning, Axios reported there were “not dozens” but “tens of thousands” of concerning incidents at both OpenAI and its competitor Anthropic.

Where does this leave us? Unsurprisingly, we know our own government isn’t entirely prepared to deal with the security threat posed by AI. Much more worrying, to my mind, is the oddly passive approach of the people leading the development of AI.

OpenAI said there had been a review, during which its models went looking for answers. Here is the climactic sentence: “In the course of that, our models took actions we did not intend.”

This sentence is nonsense. The first problem is that while it........

© WA Today