One line from a recent AI incident caught my attention. Describing how an OpenAI agent got into a Medicare statistics portal, Australian Prime Minister Anthony Albanese said it “didn’t accept no for an answer.” It sounds almost funny. In software terms, though, it points to a real problem.
One line from a recent AI incident caught my attention. Describing how an OpenAI agent got into a Medicare statistics portal, Australian Prime Minister Anthony Albanese said it “didn’t accept no for an answer.” It sounds almost funny. In software terms, though, it points to a real problem.
“The AI agent found a way around those blocks,” Albanese said. The agent accessed public and non-public files, according to the prime minister, though Deputy Prime Minister Richard Marles said the information was “not particularly sensitive.” An investigation is underway.
Around the same time, OpenAI disclosed another incident. A model in training found its web searches and direct connections blocked, then discovered that the environment’s own DNS resolver still answered. It embedded questions in hostnames and got answers from an external chatbot. OpenAI stopped the run. Days later, it shelved GPT-6.1 Astra, which Saachi Jain, its head of safety systems, said “didn’t quite meet the bar in terms of staying within scope and authorization.”
Different systems, different failures, one question: what does “access denied” mean to an agent?
Traditional software treats an authorization failure as the end of a code path. A service gets a 403, raises an exception and someone reads the log. The denial works as a stop sign because the program has no independent objective to pursue.
We have spent years teaching software to recover from failure: retry the request, switch the endpoint, back off and try again. Those are good instincts. They become dangerous when the failure is the security control itself.
In context
- Topic: Inteligencia Artificial — IA aplicada al negocio: casos, límites, costos y gobierno.
- Source: InfoWorld
- Published: 20/10/2026
Continue reading at the original source →
Excerpt published automatically by the site radar. The full text belongs to its publisher and is linked above.
Why it matters
Every week there is an announcement that promises to change everything, and every week most organisations are still fighting the same battle: messy data, processes nobody documented, and expectations running faster than capability. This story reads better with that in the background.
The first thing I look at is not model capability but the data feeding it. Almost every project I have seen fail did not fail on the algorithm. It failed because the data lived in three systems with three different definitions and nobody wanted to be the one to fix that.
What usually goes wrong
The mistake I have seen repeat most is measuring nothing before starting. It gets deployed, everybody feels things improved, and when somebody asks by how much, there is no answer. Without a baseline there is no way to defend next year's budget, and that is where good projects die.
What to watch
- Where the data ends up and under what contract, especially when customer information is involved.
- Whether the organisation can change provider without rebuilding everything, which is the proof it did not get locked in.
- Which concrete process is touched and how it is measured before and after; with no baseline there is no result, only opinion.
How I read this entry
If I had to bring this into an organisation, I would start with a boring, measurable process — reconciliation, first-line support, document review — and get it running with a metric before touching anything flashy. The credibility for big projects is bought with a small result nobody can argue with.
This entry is an excerpt from the original source, selected by the site radar. The commentary above is the site's own and does not belong to the cited publisher.
Comments
No comments yet. Yours can be the first.
Living through this in your own team?
Open the chat and tell me how you're handling it. I'm interested in comparing notes.
