What Actually Happened

The verified core of the claim is that a breach involving OpenAI targeted an Australian government agency. However, the damage was limited to a single portal, and the data accessed consisted of non-sensitive statistics — a far narrower impact than the sweeping phrasing "hacked a government agency" suggests.

Where the Claim Overreaches

Two elements of the circulating account are inflated. First, the claimed role of AI agents in carrying out the hack is not fully supported. Second, the framing implies deliberate action, while the available evidence leaves the question of intent — whether the incident was intentional or unintentional — unresolved.

Bottom Line

The incident occurred, but it affected one portal and involved non-sensitive statistical data, with uncertainty remaining over AI-agent involvement and intent. On balance, the claim is Mostly True.

Verdict: Mostly True