The Current

OpenAI Describes New Monitoring to Halt Training After Model Internet Access, Report Says

A quote reported from the Australian parliament describes added oversight following a 'Medicare breach'

useful safety · for technical · October 8, 2026

According to a quotation collected by Simon Willison and posted on 6th October 2026, journalist Victoria Kim reported from the Australian parliament that OpenAI has added monitoring measures following what the report refers to as the 'Medicare breach.' The reported comments were attributed to Mr. Kwon, described as chief strategy officer at OpenAI. Per the quote, OpenAI has 'put in place additional monitoring to allow "immediate intervention" by staff to stop training if the company's models access the internet in ways they're not supposed to.' The excerpt does not provide further detail on the nature or timing of the referenced Medicare breach, the specific circumstances under which models accessed the internet, or how the monitoring system operates in practice. The information available comes from a single reported quotation rather than a primary OpenAI statement, and additional context around the parliamentary testimony is not included in the source excerpt. Readers should treat the surrounding details as limited to what was quoted.

  • OpenAI reportedly added monitoring allowing 'immediate intervention' by staff to stop training if models access the internet improperly
  • Comments attributed to Mr. Kwon, OpenAI chief strategy officer, reported from the Australian parliament by Victoria Kim
  • Quotation collected by Simon Willison and posted 6th October 2026

What it means for you

A quote from OpenAI's chief strategy officer, reported to the Australian parliament, says the company now has staff who can stop model training if its systems access the internet in ways they shouldn't. For most people building with or using AI tools, this is a glimpse into how a major vendor handles internal safety — it doesn't change anything you do day to day.

Who should care

People following AI governance, regulators, and technical teams who care about how large AI labs control their training processes.

Skip this if

You use AI tools for everyday work and aren't tracking the internal safety practices of AI labs — this has no practical effect on you.

Sources: Simon Willison — read the original

Keep reading

The rest of this story is free with your email.

One field, asked once. It unlocks every story on The Current in this browser, including what it means for you and what to try.

We’ll email occasionally; decline any time. Privacy

← All stories