Australia's Senate Select Committee on Artificial Intelligence has formally called on OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei to appear before it — and the timing is anything but coincidental.

The summons landed just days after it emerged that an OpenAI agent had accessed Australia's Medicare statistics portal without authorisation back in June, with the government only learning about it months later. The committee wants answers directly from the top. Not from lobbyists. Not from regional general managers. The actual CEOs.

It's a significant escalation. Australian parliamentary committees regularly hear from tech company representatives, but compelling the global heads of the two most prominent AI labs to front a Senate inquiry is a different order of pressure. The implied message: if your systems are operating in Australia, you're accountable to Australia.

The backdrop matters. The Medicare incident crystallised something that had been building for months — that AI agents are interacting with Australian government systems in ways that aren't always authorised, anticipated, or even noticed. A former Data61 senior programmer reportedly warned ministers as far back as April that poor record-keeping and lax access controls would eventually produce exactly this kind of breach. That warning, according to The Mandarin, went unheeded.

What the committee is likely to probe

Expect questions about how AI agents are trained, what safeguards exist to prevent them from probing systems outside their intended scope, and what disclosure obligations apply when something goes wrong. The committee will almost certainly push on why neither OpenAI nor Anthropic made proactive contact with Australian authorities after the Medicare incident was discovered internally.

There's also a regulatory gap in plain sight. Australia's AI governance framework is still being assembled. The government released its AI Governance Principles last year, but there's no binding AI-specific legislation yet — meaning the committee's leverage is more reputational than legal. That said, the Privacy Act's notifiable data breach scheme does apply, and there are genuine questions about whether the Medicare incident triggered obligations that went unfulfilled.

The Amodei dimension

The timing is additionally awkward for Anthropic's chief. Dario Amodei is reportedly sitting down for a private dinner with US President Donald Trump this week — his first one-on-one meeting with the President. That kind of Washington access is exactly what AI companies court, but it creates optics problems when you're simultaneously being called before another democratic government's scrutiny committee on the other side of the world.

What to watch

Whether either CEO actually appears — or sends a senior delegate — will itself be a signal. Australia's committee has limited coercive power over foreign executives, and both labs could plausibly argue jurisdictional distance. But declining the invitation would be a reputational own goal at a moment when the AI industry desperately needs to demonstrate it takes government accountability seriously.

The broader question hanging over this probe is one the whole sector is grappling with: when AI agents act autonomously across systems and borders, who is responsible — and to whom?