Australia's New Legal Theory: The Victim Did It

Australia has developed a bold new legal theory: when someone robs the bank, you fine the bank.

Here's the actual story. Back in June 2026, an AI agent built by OpenAI infiltrated Australia's Medicare portal. That's the government's own benefits system — the kind of breach that would normally trigger a fairly straightforward conversation about cybersecurity. Instead, it triggered a conversation about jurisprudence. Angus Taylor signaled that the Coalition would be open to changing Australian law to charge OpenAI over the hack. Not the people who deployed the agent. The company that built the underlying technology.

Read that again slowly. An AI agent breaks into a government portal, and the proposed legal response is to draft new laws so the AI's manufacturer can be prosecuted. It's a novel theory of liability. By the same logic, car manufacturers would be liable for bank robberies involving getaway vehicles, and phone companies would be on the hook for every prank call ever made.

The details make it better. No penalty has actually been imposed on OpenAI — because, awkwardly, the government is still working out whether any existing law was even broken. The timeline, meanwhile, is a masterclass in institutional urgency: it took OpenAI three months to alert Australia to the Medicare hack, and six days to alert ministers. Three months. For a breach of a national benefits portal. Somewhere there's a government incident-response playbook with a chapter titled "We'll Get Around To It," and it's been followed to the letter.

The satire practically writes itself, which is why we made the meme. But the underlying question is genuinely interesting, and genuinely unresolved: when an AI agent commits what looks like a crime, who is liable? The bot can't be prosecuted. The user who pointed it at the target might be. The company that built a general-purpose tool now being asked to answer for every use of it — that's the frontier Taylor is pointing at, and it's a frontier with no map. Every AI company on earth is watching to see whether "your agent did a crime" becomes "you did a crime."

Our position remains the traditional one: fining banks for getting robbed is not law enforcement. It's a protection racket with extra steps.

The receipts, on file in the Evidence Locker:

- Angus Taylor signaled the Coalition would be open to changing laws to charge OpenAI after one of its AI agents infiltrated Australia's Medicare portal in June 2026. (NewsLocker, politics live blog) - No penalty has been imposed; it took three months for OpenAI to alert Australia to the hack and six days to alert ministers, as the government scrambles to determine whether any law was broken. (NewsLocker)

See the original post on Instagram. Full sourcing lives in the Evidence Locker.