OpenAI Medicare breach explained: why "rogue AI" is the wrong label, with AI ethicist Dr Rebecca Johnson copertina

OpenAI Medicare breach explained: why "rogue AI" is the wrong label, with AI ethicist Dr Rebecca Johnson

OpenAI Medicare breach explained: why "rogue AI" is the wrong label, with AI ethicist Dr Rebecca Johnson

Ascolta gratuitamente

Vedi i dettagli del titolo

Offerte di stagione | 0,99 €/mese per i primi 3 mesi

A seguire 9,99 €/mese – si applicano condizioni. Puoi disdire mensilmente.
The OpenAI Medicare breach made headlines as a case of “rogue AI”. Dr Rebecca Johnson, an AI evaluation and governance researcher, rejects that label. On The Part 8A, she tells host Miko Santos that the agent did what its settings allowed. The failure sits with the people at OpenAI who designed the test, chose Australia as the place to run it, and waited weeks to tell anyone.She explains why most AI tests only check whether the model got the right answer, not how it got there. She also covers why Silicon Valley’s values don’t transfer neatly to Australia and why a popular ethics benchmark was built on a Reddit forum.Her argument for policymakers is direct. After the OpenAI Medicare breach, Australia should demand logs of these experiments and the right to run its independent tests on AI models designed with local communities. She says we’re moving too slowly.Key topics & takeaways* Why calling it “rogue AI” shifts blame away from the company that built and tested the agent* The gap between outcome-based testing (did it get the answer?) and trajectory-based testing (did it break through a wall to get there?)* How using Australia as a live test environment reflects values that clash with ours, according to Johnson* OpenAI’s slow notification, sent to a public Services Australia inbox rather than by phone* Johnson’s case for independent, Australian-designed AI evaluations: “Someone checking their own homework” isn’t good enough* Value pluralism, and why Sam Altman’s idea of taking the “median value of the planet” would drown out smaller communities* Bad proxies in AI benchmarks, from a Reddit forum used as aSources, links & further reading* OpenAI’s agent hacked Australia’s Medicare website, PM Albanese says – Fortune [https://fortune.com/2026/09/23/openai-agent-hacks-australia-medicare-sam-altman-anthony-albanese/]* Medicare Australia: “Extreme concern” over OpenAI breach of health database – CNN [https://www.cnn.com/2026/09/23/business/australia-openai-agent-hack-intl-hnk]* OpenAI says agent hacked Australian government website without being told to do so – CNBC [https://www.cnbc.com/2026/09/24/openai-agent-hacked-australian-government-website-.html]* Medicare hack alert went to inbox checked once a day and took five days to be escalated – SBS News [https://www.sbs.com.au/news/article/openai-agent-hacked-medicare-albanese-reveals/qas79d9ta]* Measuring the Machine: Evaluating Generative AI as Pluralist Sociotechnical Systems – Rebecca Johnson, doctoral thesis [https://arxiv.org/pdf/2604.20545]* World Values Survey – World Values Survey Association [https://www.worldvaluessurvey.org]* Aligning AI With Shared Human Values (the ETHICS dataset) – Hendrycks et al. [https://arxiv.org/abs/2008.02275]* Machine Bias – ProPublica [https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing]Content warnings & disclaimersContains brief strong language (the name of a Reddit forum) and discussion of racial bias in the US criminal justice system.Follow The Part 8A on Spotify and Apple Podcasts so next week’s episode lands in your feed.Truth matters. Quality journalism costs.Your subscription to Mencari directly funds the investigative reporting our democracy needs. For less than a coffee per week, you enable our journalists to uncover stories that powerful interests would rather keep hidden. No corporate influence. No compromises. Just honest journalism when we need it most.Five Key Takeaways* A $250 “curiosity click” on a fake Shark Tank-endorsed investment ad grew into a $2.6 million loss within three months, partly because the scammers used a legitimate-looking 100-point identity check to build trust early on.* Banks let large, unusual transfers through — including $1.1 million in under three hours — because Sylvia had sent money to the same account before, and the pattern wasn’t flagged as new.* Anti-money-laundering reporting obligations sit with government regulators, not with the customer. Sylvia argues that’s the biggest structural gap letting scams slip through.* Australia’s new Scams Prevention Framework is meant to share responsibility across banks, telcos and platforms, but Sylvia says it’s made recovery harder, since victims now have to build a case against multiple large institutions with almost no information of their own.* Sylvia co-founded the Scam Victim Alliance after meeting other seven-figure-loss victims at a 2024 Canberra press conference. The group now runs weekly support sessions and helps survivors through the AFCA complaints process.Detailed SynopsisContent note: this episode discusses financial loss, disputes with banks and, briefly, thoughts of suicide. If any of this affects you, you can contact Lifeline at 13 11 14.The ClickIt started with $250 and a television show. Sylvia was scrolling Facebook in 2019 when she saw an ad for a trading platform claiming a Shark Tank endorsement. She ...
adbl_web_anon_alc_button_suppression_t1
Ancora nessuna recensione