top of page

AI Agent Incursions: Lessons for Companies from Recent Government Experiences

28 Oct 2026

October is Cyber Security Action Month in Australia, and this year’s theme from the Australian Signals Directorate is “Take a second.”  It’s a good moment to take a second over a recent event which should change how every organisation thinks about data protection.


The incident that changed the conversation

On 24 September, the Australian Government revealed that an OpenAI agent had gained unauthorised access to files on a government health data portal, the Medicare Statistics Reporting Service. It may be the first known case of an AI agent breaching a government website. The agent got in on 18 June, but OpenAI didn’t disclose it until 10 September.


It didn’t stop there. An OpenAI agent was also reported to have accessed a NSW National Parks web app, an incident classified as a “misalignment”. Home Affairs has since ordered a government-wide stocktake of legacy systems within six months.


Why it matters for enterprises

It’s tempting to file these as government problems, or as a story about one AI vendor. However, that would be a mistake, for three reasons.


1. Intent no longer defines the threat.

These agents weren’t built to attack anyone. The agent bypassed access blocks for ordinary research purposes. An autonomous system that treats a security control as an obstacle to route around is a risk whatever its creator intended. Any AI agent operating in or near your environment - whether yours, a vendor’s or a customer’s - can do the same.


2. Your exposure is wider than you think.

Agents browse, authenticate, call APIs and follow links at machine speed. Any web app, portal or file share that relies on obscurity, a weak access control or a forgotten configuration is now reachable by something that never tires and never gets bored. Legacy systems, the focus of the new government stocktake, are the obvious weak points.


3. Disclosure gaps compound the damage.

Nearly three months passed between this incident and its public disclosure. In your organisation, that kind of delay means customers, regulators and partners learn about a problem long after the opportunity to contain it has gone.


The cost of losing customer and enterprise data is already well documented in Australia. Origin Energy reported that a late-July breach exposed credit card and bank account details of around 900,000 current and former customers. Optus, Medibank and Qantas have shown how far the consequences extend: regulatory scrutiny, class actions, lost trust and enormous remediation costs.


AI agents add a new way for data to leave the building, and it’s one most security programs weren’t designed for.


What should companies do?

Start by knowing where your sensitive data lives and how it moves. Then control that movement. Much of the data that gets lost is in transit: files exchanged with partners, customers and systems through ad hoc scripts, shared drives, email attachments and unmanaged web portals. That’s where AI agents, and the people and tools behind them, find easy openings.


How a well-configured MFT helps

A modern managed file transfer (MFT) solution such as GoAnywhere MFT gives you a controlled, auditable path for data movement in place of a sprawl of informal ones. This helps in several ways:


  • Least-privilege access. Every user, system and agent gets only the access it needs, with strong authentication, multi-factor where appropriate, and IP restrictions.


  • No open doors. Data moves through an encrypted, authenticated channel instead of sitting on exposed web apps or file shares an agent can wander into.


  • Visibility. Full audit trails and alerting show who or what accessed which file, and when, so you can spot unusual behaviour in hours rather than months.


  • Automation with guardrails. Governed, automated workflows replace ad hoc scripts and the temptation to hand an agent broad access to get a job done.


  • Layered defences. Encryption, antivirus and data loss prevention integration add checks that a determined agent can’t simply bypass.


An MFT narrows the number of places your data can leak from, and it gives you the evidence to respond quickly when something goes wrong.


Take a Second

This Cyber Security Action Month, take a second to ask: if an AI agent went looking for your customer data today, what would it find, and who would know?


If you’d like to talk through how to tighten up your data transfers, we’d be glad to help at Generic Systems Australia.  We have more than a decade of experiencing helping Australian and New Zealand customers install and leverage the world’s leading MFT.

bottom of page