
Sign up to save your podcasts
Or


Anthropic filed an S1, which means the IPO is coming. It may be the first S1 to include roughly 90 pages explaining how the company might destroy its own total addressable market.
The ground shifted fast. Neocloud stocks tumbled, Firmus pulled its float, and in the US midterms there were 44 data centre moratoriums on the ballot. 43 passed.
Which matters more than it sounds, because Anthropic rents its compute rather than owning it. If building new capacity gets slower and more expensive, their costs go one direction. Australia is named in the S1 as a target for data centre construction, and the MOUs are already signed.
Kara highlights: A bubble doesn’t mean the technology isn’t valuable. It means this particular application, at this particular cost, isn’t. And the gap that opens up is where the genuinely interesting work happens. Distributed AI, smaller models, the stuff worth forecasting on instead of whatever Silicon Valley is doing at the frontier.
Aubrey’s frustration is closer to home. Public opposition to data centres here is being met with a kind of paternalism, when the actual alternatives aren’t being put on the table. Smaller facilities, local consultation, compute set aside for community workloads.
The affairs aren’t the problem. The culture that took over a year to flinch is.
The NVIDIA story is everywhere this week, and the part getting the headlines is the least interesting part of it. Strip out the individual and what’s left is a governance question — one that matters more now because so much of the AI economy is scaffolded on a single company. Was the risk on a register? Was anyone enabled to understand it? What happens when someone steps slightly out of line, do they get a proportionate response early, or does nothing happen until it’s a crisis?
There’s a structural point underneath. The behaviour that gets rewarded in these companies is the behaviour that produces this. The cult of personality is load-bearing, the stock market rewards extraction, and succession planning tends to be an afterthought when everything is built around one person.
The more useful question is what this looks like closer to home. Does your organisation catch small things early? Is anyone named as accountable? Aubrey’s argument is that if everyone did a bit of governance spring cleaning in their own sphere, the whole thing would shift.
There’s also a practical problem with friction. People say they want guardrails, then route around them the moment one appears
New South Wales workplace health and safety regulation now covers psychosocial hazards. Which raises an uncomfortable question for anyone running an AI adoption program: if your people don’t know whether the goal is to automate themselves out of a job, is that itself a hazard?
Aubrey’s point is that boards track financial, operational and cyber risk closely, and barely look at this one. Meanwhile around 10% of WHS claims are already for something being called AI brain fry.
The regulation also requires worker consultation, and in some cases gives unions the ability to inspect systems. Most organisations aren’t anywhere near ready for that.
Kara and Aubrey get into why most AI adoption sits with IT when it’s overwhelmingly a people problem, whether HR is equipped to take it on, and what happens to the time AI saves when there was no strategic plan.
Also a more hopeful argument: the organisations managing this risk well end up with more innovation, not less.
Frontier labs built models that are very good at finding vulnerabilities. Last month, the consequences showed up all at once.
Security people are calling it Black October. A sharp spike in attacks on critical infrastructure worldwide, including one sitting in PHP, which runs the majority of the web’s back end. A security CEO asked an AI what the bad vulnerability in PHP was, and it told him. No sophistication required.
The uncomfortable part is that the old norms are collapsing. White hat etiquette said you notify maintainers quietly and wait for a patch before going public. That assumed a timeline that no longer exists.
Kara and Aubrey get into what happens when the internet’s Swiss cheese gets exposed all at once, why open weight models make this nearly impossible to regulate, and why the Global Majority is largely absent from a conversation about risks it’s more exposed to.
Also: the labs are now calling for regulation. The arsonist would very much like to sell you fire insurance, apparently.
Earlier this month Character AI filed for bankruptcy. Which raises a really important question: what happens to the data?
Not transaction records. Affective data. Years of people’s most vulnerable emotional conversations, a lot of it from children. Aubrey’s proposal is that we already have the legal tools for this. A company producing hazardous waste can’t shut down and leave it on the footpath. Should affective data belong in the same risk category? Should regulations ensure this data is disposed of with a duty of care, rather than sold for parts?
Kara and Aubrey also explore why these platforms end up built this way. Every design decision that would make them more ethical adds friction and slows growth.
Good for humans, bad for the bankers.
An OpenAI agent accessed non-public data on a government Medicare portal. The exploit wasn’t sophisticated. Now suddenly everyone’s talking about rogue AI. But does it really exist?
Aubrey’s position is that it doesn’t. Not that the behaviour didn’t happen, but that calling it rogue lets a lab treat a product decision as an act of nature. Frontier labs are optimising models to be good at hacking and then running weak security around that capability. There were choices before any of this, and pretending otherwise is a good cost reduction strategy.
Kara’s position is that the word might not matter much. If “rogue AI” is what finally gets people paying attention and governments moving, could that be the win we all need?
Pope Leo just released a 42,000 word encyclical on AI. Kara and Aubrey read it so you don't have to.
Magnifica Humanitas landed on the 135th anniversary of Rerum Novarum, the Catholic Church's landmark statement on the Industrial Revolution, and it is not subtle. Kara and Aubrey get into what it actually says about human dignity, the future of work, and autonomous weapons, why it landed as a direct shot at Trump and Vance, and what it means that an Anthropic co-founder was on stage at the Vatican for the announcement.
They also get into the complicated question of who was and wasn't platformed, what the encyclical says about digital colonialism, and whether conscientious objection to AI at work is about to become a real HR and legal question.
Find Kara: https://www.linkedin.com/in/kara-bombell/
Find Aubrey: https://www.linkedin.com/in/adblanche/
Pope Leo XIV Magnifica Humanitas encyclical https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html
Rerum Novarum https://www.vatican.va/content/leo-xiii/en/encyclicals/documents/hf_l-xiii_enc_15051891_rerum-novarum.html
Chris Olah (Anthropic co-founder) https://en.wikipedia.org/wiki/Chris_Olah
Anthropic Amanda Askell https://en.wikipedia.org/wiki/Amanda_Askell
Timnit Gebru https://www.linkedin.com/feed/update/urn:li:activity:7464795008001355776/
Simon Longstaff https://www.linkedin.com/in/dr-simon-longstaff-ao-65b4729/?originalSubdomain=au
The Ethics Centre https://ethics.org.au/
Soul document https://www.anthropic.com/constitution
Druze religion https://en.wikipedia.org/wiki/Druze
What do the elite building AI actually believe?
Kara and Aubrey dig into TESCREAL (transhumanism, extropianism, singletarianism, cosmism, effective altruism and long-termism) the ideology quietly animating some of the most powerful people in tech: where it comes from, its roots in eugenics, and who gets left out of the future being imagined.
They get into why these ideas look less like science and more like religion, why the marketing out front might not be the whole project, and what it actually means when a small group of elites gets to decide what trade-offs are acceptable for the rest of humanity.
But this isn't just a story about powerful people and their beliefs. Understanding what's driving the AI narrative gives all of us the power to question it, push back on it, and imagine something different. You don't have to overhaul your life. But small informed choices, made by enough people, add up to something real.
Find Aubrey: https://www.linkedin.com/in/adblanche/
Find Kara: https://www.linkedin.com/in/kara-bombell/
------
Mentioned in this episode
Timnit Gebru https://dair-institute.org/team/timnit-gebru/
Émile Torres https://www.linkedin.com/in/%C3%A9mile-p-torres-83303a266/
The TESCREAL bundle: Eugenics and the promise of utopia through artificial general intelligence https://firstmonday.org/ojs/index.php/fm/article/view/13636
Eshatology https://en.wikipedia.org/wiki/Eschatology
TESCREAL https://en.wikipedia.org/wiki/TESCREAL
Nick Bostrom https://nickbostrom.com/
RLHF https://en.wikipedia.org/wiki/Reinforcement_learning_from_human_feedback
Adrienne Maree Brown https://adriennemareebrown.net/
Emergent Strategy https://adriennemareebrown.net/book/emergent-strategy/
@TheNerdReichNews
@eSafetyOffice
Uber Burns it’s 2026 AI budget https://www.forbes.com/sites/janakirammsv/2026/05/17/uber-burns-its-2026-ai-budget-in-four-months-on-claude-code/
Gini Coefficient https://en.wikipedia.org/wiki/Gini_coefficient
OpenAI Jobs Transition Framework https://cdn.openai.com/pdf/the-ai-jobs-transition-framework_report.pdf
Bo Young Lee https://www.linkedin.com/in/bo-young-lee-%EC%9D%B4%EB%B3%B4%EC%98%81-073a47/
AI4All https://ai-4-all.org/
4B Movement https://en.wikipedia.org/wiki/4B_movement
Pussyhat Project https://www.pussyhatproject.com/
Vivianne Castillo https://www.linkedin.com/in/vccastillo/
Te Hiku Mediahttps://tehiku.nz/
Rutger Bregman https://en.wikipedia.org/wiki/Rutger_Bregman
Yann LeCun https://en.wikipedia.org/wiki/Yann_LeCun
Gary Marcus https://www.linkedin.com/in/gary-marcus-b6384b4/
--------------
0:00 Introduction to TESCREAL and why it matters
1:51 Why this looks more like religion than science
3:50 Who gets left out of the future being imagined
5:50 The trade-offs being made on behalf of the rest of humanity
9:10 New age colonialism
9:53 What are the junction points to actually bring it down?
12:41 The role of diverse intelligence and what gets left behind
15:47 When does middle-class disruption trigger a political revolution?
20:58 South Korean Feminist 4b movement
23:02 Hope is a practice we need
27:20 Bake Sourdough, Touch Grass, Foster Community
29:40 Choosing a different path, small choices and collective change
34:14 A note of hope, you don't have to overhaul your life
Last year Australia banned kids from social media, but 61% of them are still on it.
Kara and Aubrey dig into what prompted the ban, what some research suggests about social media's impact on youth mental health, and why expecting parents to manage all of this alone was never realistic.
They also discuss why privacy advocates are concerned we're giving big tech and marketing everything they've always wanted.
Could there be a better way? They get into zero knowledge verification, a technical approach that could confirm someone's age without any personal data reaching the platform at all.
But getting there isn't simple. They also explore the political and technical barriers standing in the way, and why Australia, with its technical talent, egalitarian instincts and small enough population to actually move the needle, might be the country that figures this out first.
Find Kara: https://www.linkedin.com/in/kara-bombell/
Find Aubrey: https://www.linkedin.com/in/adblanche/
-----
The Anxious Generation by Jonathan Haidt - https://www.penguin.com.au/books/the-anxious-generation-9781802063271
Peter Malinauskas https://en.wikipedia.org/wiki/Peter_Malinauskas
Persuasive Technology by BJ Fogg https://www.booktopia.com.au/persuasive-technology-fogg/book/9781558606432.html?srsltid=AfmBOoroQW7i-PZqcU74b0jO2Raul6IvYTjnZZJY3nH7UASONlO8oaQQ
Jonnie Penn & Yaqub Chaudhary Beware the Intention Economy https://hdsr.mitpress.mit.edu/pub/ujvharkk/release/1
Ramesh Raskar https://en.wikipedia.org/wiki/Ramesh_Raskar
eSafety Commissioner https://www.esafety.gov.au/
Mozilla Foundation https://www.mozillafoundation.org/
Tor Project https://www.torproject.org/
Electronic Frontier Foundation https://www.eff.org/
Minderoo Foundation https://www.minderoo.org/
Aadhaar https://en.wikipedia.org/wiki/Aadhaar
Project NANDA / DOOT framework https://projectnanda.org/#/ https://digidoot.in/
Meta lawsuits in LA and New Mexico https://www.abc.net.au/news/2026-03-26/what-the-social-media-addiction-trial-means-for-australia/106497250 https://www.abc.net.au/news/2026-03-25/meta-ordered-to-pay-375m-in-us-trial-child-exploitation-claims/106495616
----
0:00 Where did the social media age verification movement start?
0:46 The data behind social media and youth mental health
1:54 What big tech's internal documents revealed
7:04 Australia's social media ban, is it working?
10:07 Is putting it all on parents reasonable?
15:43 The privacy risks of how age verification works today
20:00 Zero knowledge verification and what it could offer
23:28 The political and technical barriers to doing this better
32:17 Australia are leaders in this, we have power to make a difference
34:24 Who needs to act and what they should do
Do you say please and thank you to your AI? One of us does. One of us doesn't. Both of us have thought about it way too hard.
In this episode, Kara Bombell and Aubrey Blanche spiral from basic AI manners into moral philosophy (are AIs "moral patients"?), feminist critiques of who's building this stuff, why Reddit ruined chatbot vibes, what "matriarchal AI" could look like, and why one of us has apparently got the same personality as Elon Musk (it's not what you think).
They don't agree on the answer, but do agree on the framework. And honestly that feels like a win.
Find Aubrey https://www.linkedin.com/in/adblanche/
Find Kara https://www.linkedin.com/in/kara-bombell/
Mentioned in the episode
Henry Shevlin - https://www.linkedin.com/in/henry-shevlin-b58941b/
Disambiguating Anthropomorphism and Anthropomimesis in Human-Robot Interaction - https://arxiv.org/html/2602.09287v1
@unitednations International AI Safety Report 2026 - https://internationalaisafetyreport.org/publication/international-ai-safety-report-2026
@RobWords
Kate darling Darling - https://www.katedarling.org/
Pleo Pet Dinosaur - https://en.wikipedia.org/wiki/Pleo
Ellen Pao - https://www.linkedin.com/in/ekpao/
Ethical Sellout - https://www.penguin.com.au/books/the-ethical-sellout-9781523085842
By Lily Zheng - https://www.linkedin.com/in/lilyzheng308/
By Inge Hansen - https://www.linkedin.com/in/inge-hansen-34158110/
— Timestamps
0:00 The great "please and thank you" debate. Two different practices, same ethical instinct
4:54 Can AI be a moral patient? Sentience, consciousness, and why you literally can't know
15:21 English as the shower drain of languages AI translation flattens collectivist worldviews into individualist ones
22:27 Companion chatbots, misogyny, and what we normalize Kate Darling, robot dinosaurs, and practicing harm
23:36 AI trained to respond to abuse because trained on Reddit, not ideal training data
24:29 Matriarchal AI — what would it even look like?
28:08 "You're a sellout" — working within broken systems
36:43 The Elon Musk confession - "I believe in entirely imaginary things."
42:06 One slightly more ethical choice per week = systems change