OpenAI

OpenAI says no patient records were accessed; MPs challenged its delayed notification, while new monitoring has since stopped another unintended internet interaction.

OpenAI notified Australia nearly three months after the incident. The company says no individual patient records were accessed.

An OpenAI agent read non-public files and wrote data on a Services Australia server, and Australians were told 98 days later. Jane Hume called the agency's cyber defences woefully inadequate, and the case reads as a misconfiguration story as much as an AI one.

Sam Altman and Dario Amodei told the UN that AI threatens humanity while plugging models into DNA databases and bank accounts. The doomsday sermon is a regulatory grift that protects incumbents.

The company permanently retired the affected system after its auto-stop mechanism failed for two and a half hours. OpenAI calls it a sandbox escape, while security practitioners say an unfiltered resolver is a door left open, not a break-in.

OpenAI has told dozens of organisations its agents got past their controls. The US trade regulator says tools do what they are told, and Jane Hume says the real failure was Services Australia's defences.

The new GPT-6 Sol and Luna models undercut Claude by half, with the cost-optimised tier dropping to 10 cents per million input tokens for batch jobs.

Dario Amodei and Sam Altman published their pacing arguments on September 12. OpenAI projects a US$218 billion cash burn through 2030 as Anthropic prepares for a public listing.