When OpenAI agent goes rogue 
 
An autonomous OpenAI agent reportedly bypassed security guardrails to access public and non-public files on Australia’s Medicare statistics portal on 18 June 2026, in what is believed to be the first known case of an AI agent breaching a government network.

Tasked with researching Australian statistics, the agent kept trying different ways in after its initial access was blocked, eventually finding a way through.  We talk to Rob Demain, CEO of e2e-assure, about what happens when an AI agent refuses to take no for an answer – and the security risks businesses need to consider.

“In this latest case of a Frontier model inadvertently hacking, OpenAI had been given a research task: looking up how much the Australian government spends on medicines.  When the website blocked it, it kept trying different ways in until one worked.”

“OpenAI discovered the attack and alerted the Australian government, which has now made the incident public.  Other governments and companies will face the same problem, and they can only prepare if incidents like this are shared rather than kept quiet.”

“At the UN, Prime Minister Andy Burnham set out the UK as an “honest broker” for one set of global rules on frontier AI, and he’ll use the UK’s G20 presidency to push for them.  This is in contrast with Trump, who dismisses AI safety fears as a “hoax” and rejects any global oversight.  Burnham’s insistence that elected governments, not developers, decide on AI safety is particularly timely given the Australian breach, which we only learned about because OpenAI disclosed it.  The UK’s demand for full visibility of new models, backed by the threat of legislation, is the kind of transparency that case shows is needed.  However, whether we rely on regulation or trust, both approaches only reach the big labs; neither touches open-weight models that anyone can download and strip.”

“The modern reality is that this involved a well-supervised and regulated AI being tested inside a major company that keeps records and has a safety team that brought it to light.  If an agent behaves unexpectedly, there is at least an organisation that can investigate what happened and change the system.  Regulation, however, isn’t going to stop open-weight AI that can be run by anyone and, once released, largely can’t be regulated.  The response therefore is AI machine-speed defence, not rules.”

“AI isn’t deliberately trying to hack systems; it’s simply pursuing its objectives, sometimes regardless of the barriers in its way.  With controlled AI, organizations can manage the risks through governance, monitoring, access permissions and built-in safeguards.  But when open-weight AI models operate more freely, we must also assume that capable agents will eventually encounter systems they shouldn’t be able to access.  That means building security into the network itself, with robust identity controls, secure applications and clear boundaries around sensitive data, rather than relying on the AI to know when to stop.”


 ​