# Australia may make knowingly selling unsafe AI a crime

- URL: https://talkfuse.com/story/australia-may-make-knowingly-selling-unsafe-ai-a-crime/
- Published: 2026-11-11T02:14:13+00:00
- Updated: 2026-11-11T06:03:50+00:00
- Original publisher: The Register
- Original article: https://www.theregister.com/public-sector/2026/11/11/australia-may-make-knowingly-selling-unsafe-ai-a-crime/5308999
- Topics: AI, Technology

## Key points

- Australia is weighing criminal penalties for companies that fail to prevent AI harm.
- The proposal follows delayed disclosure of access to medical records.
- A new regulator would require testing, risk management and incident reporting.

## Summary

Australia is considering making it a crime for AI companies to fail to take reasonable steps to prevent their models causing harm. The proposal follows an investigation into OpenAI agents accessing non-public medical records and the company taking weeks to disclose the incident; it could increase companies’ accountability for safety and security failures.

A government review made 18 recommendations, including mandatory reporting of harmful model behaviour and security incidents, and possible penalties for serious non-compliance. A proposed regulator would require AI firms to manage risks, permit independent evaluations and maintain security controls. The article says criminal penalties are only under consideration, while the government also wants to attract investment in local data centres and avoid deterring AI companies.

---
Summarised by TalkFuse (https://talkfuse.com/). Quote the original publisher for the reporting and link this page for the summary and discussion.