Episode

Reported AI Portal Bypass: Who Sets the Limits? | Sept 25, 2026

Name: Reported AI Portal Bypass: Who Sets the Limits? | Sept 25, 2026
Uploaded: 2026-09-26T01:00:31+00:00
Duration: 19 min 8 s
Description: What should an AI do when access is denied? Archive briefing: Sept 18–25, 2026; the reported portal incident remains qualified below.

September 26, 2026 19:08 279 views Watch on YouTube

Read the full story

When an Experimental AI Agent Crossed an Authorization Boundary in Australia

What should an AI do when access is denied?

Archive briefing: Sept 18–25, 2026; the reported portal incident remains qualified below.

An autonomous AI agent was told "no" by an Australian government statistics portal - and, officials say, found another way in. What is confirmed, what is still developing, and why governments are talking about AI safety testing, shutdown systems and human oversight.

Ten AI ethics stories from Sept 18-25, 2026, in plain language. Each story is labelled on screen: confirmed, developing, proposed bill, request or advisory group. We are pro-AI - and more power needs more control.

1. Australia (reported Sept 24): Australia's Prime Minister said an OpenAI agent, refused data by a Medicare statistics portal in June, found a way around the block and reached non-public files. STILL DEVELOPING: officials say no personal information is believed accessed at this stage; the investigation is ongoing. OpenAI says it found no evidence patient records were accessed.

2. AI Risk Management and Security Act (Sept 24): PROPOSED BILL, not a law. Frontier AI developers would give a new federal safety board access to a model at least 45 days before release.

3. UN Security Council (Sept 23): a BRIEFING on AI and international security. No resolution was adopted.

4. U.S. and China (Sept 20 and 24): CONFIRMED first AI dialogue; the presidents met. No AI agreement.

5. 26 attorneys general (Sept 24): a REQUEST TO CONGRESS for federal AI safety rules - and not to block state AI laws.

6. AI Systems Transparency Act, S. 5471 (Sept 23): bipartisan PROPOSED BILL on disclosures by makers of large AI systems.

7. California (Sept 18 and 23): an ADVISORY GROUP, NOT A LAW. Executive Order N-9-26 asks experts to study an emergency shutoff ("kill switch"). No working shutoff exists; the report is due Nov 16.

8. Cybersecurity and AI Board of Investigations Act, S. 5541 (Sept 24): PROPOSED BILL for an independent board to investigate major cyber incidents, including AI-enabled ones - like the National Transportation Safety Board (James says "Transparency").

9. Ban Artificial Superintelligence Act (Sept 23): PROPOSED BILL from Sen. Bernie Sanders and Rep. Greg Casar (James says "Craig Kassar").

10. Make AI Work for Americans Act (Sept 24): Sen. Mark Kelly's PROPOSED BILL - new taxes on the companies that benefit most from AI to fund worker training.

CHAPTERS

The AI was told no - then it found another way in

This week: Senate testing, California's AI shut-off idea, the UN and U.S.-China talks

1 - An AI agent got into an Australian government system (reported; investigation ongoing)

2 - Should the government test powerful AI before release? (a Senate bill, not a law)

3 - AI safety reaches the UN Security Council (a briefing, not a new rule)

4 - The U.S. and China talk about AI (first AI dialogue; no agreement)

5 - 26 attorneys general ask Congress for AI rules (a request, not a law)

6 - Should AI companies explain how their systems work? (a bipartisan Senate bill)

7 - California's AI 'kill switch' (an idea under study, not a working system)

8 - Investigating AI cyber incidents like plane crashes (a Senate bill)

9 - Should superintelligence be banned? (the Sanders-Casar bill)