# A big-tech defector’s AI warning for the world

An episode of [7am](https://invisiblenetwork.ai/podcast/1461999702), 15 Sep 2026, 18 min.

Audio: https://p.podderapp.com/2544644999/traffic.omny.fm/d/clips/2fb3740d-3436-44af-8cc0-a91900716aa5/62723a1f-3e96-4fc4-882c-b25e0181da4d/e26765a1-2f5e-4eca-880f-b4c60055acfc/audio.mp3?utm_source=Podcast&in_playlist=70f29d90-2a63-48aa-bf8f-b25e0181da70

In the last week, AI bosses around the world have suddenly grown a conscience – urging a global slowdown amid fears the race to develop artificial intelligence could outstrip humanity’s ability to control it. For Dex Hunter-Torricke, that realisation came a little sooner. The big-tech insider walked away last year, disillusioned, after spending 15 years inside the belly of the beast. He wrote speeches for Mark Zuckerberg at Facebook, worked with Elon Musk at SpaceX, and most recently helped Google DeepMind explain the rise of artificial intelligence to the world. Now, he says society is making the same mistakes with AI as it did in the social media era – sleepwalking into what comes next. Today, founder and president of The Center for Tomorrow, Dex Hunter-Torricke gives us a glimpse inside the big tech firms shaping our future. If you enjoy 7am, the best way you can support us is by making a contribution at 7ampodcast.com.au/support. Socials: Stay in touch with us on Instagram Guest: Founder of The Center for Tomorrow Dex Hunter-Torricke Photo: @dexhuntertorricke via Instagram See omnystudio.com/listener for privacy information.

## Transcript

[0:01] Speaker 1: I'm Daniel James, and you're listening to 7am. In the
[0:09] Speaker 1: last week, AI bosses around the world have suddenly grown
[0:13] Speaker 1: a conscience, urging a global slowdown amid fears the race
[0:16] Speaker 1: to develop artificial intelligence could outstrip humanity's ability to control it.
[0:22] Speaker 1: For Dex Hunter Torek, that realisation came a little sooner.
[0:25] Speaker 1: The big tech insider walked away last year. disillusioned after
[0:29] Speaker 1: spending 15 years inside the belly of the beast. He
[0:32] Speaker 1: wrote speeches for Mark Zuckerberg at Facebook, worked with Elon
[0:36] Speaker 1: Musk at SpaceX, and most recently helped Google DeepMind explain
[0:40] Speaker 1: the rise of artificial intelligence to the world. Now he
[0:44] Speaker 1: says society is making the same mistakes with AI as
[0:47] Speaker 1: it did in the social media era, sleepwalking into what
[0:50] Speaker 1: comes next. Today, founder and president of the Centre for Tomorrow,
[0:54] Speaker 1: Dex Hunter-Torik. gives us a glimpse inside the big tech
[0:58] Speaker 1: firms shaping our future. It's Wednesday, September 16. Dex, thanks
[1:12] Speaker 1: so much for joining us. You spent 15 years helping
[1:15] Speaker 1: some of the most powerful people in tech sell the
[1:18] Speaker 1: promise of technology. When did you stop believing technology?
[1:22] Speaker 2: I've never actually stopped believing in the promise of technology.
[1:26] Speaker 2: Technology can and should be something deeply hopeful. What I did, though,
[1:31] Speaker 2: was I stopped believing that a bunch of those leaders,
[1:35] Speaker 2: a bunch of those companies are configured to really serve
[1:39] Speaker 2: the interests of most people in our societies or around
[1:42] Speaker 2: the world. I think, you know, connecting people, that is
[1:46] Speaker 2: a good thing, but you have to manage it well.
[1:49] Speaker 2: You get a lot of new challenges that come along
[1:51] Speaker 2: with building this totally new form of human community in
[1:55] Speaker 2: the online realm in a world of eight and a
[1:57] Speaker 2: half billion people at this moment in history. And it
[2:01] Speaker 2: puts responsibilities on platforms to ensure that they're meeting their
[2:05] Speaker 2: obligations to protect users, that they're really providing the kind
[2:09] Speaker 2: of ecosystem in which healthy discourse can unfold. And I
[2:13] Speaker 2: think after 2016, when we saw the arrival of Donald
[2:18] Speaker 2: Trump as president, we saw increasing political populism and volatility.
[2:24] Speaker 2: We saw this huge spike in disinformation, hate speech online.
[2:28] Speaker 2: And I went on to work after Meta for an
[2:31] Speaker 2: organization called the Oversight Board. which was a body that
[2:35] Speaker 2: was set up to actually scrutinize content decisions on Facebook
[2:38] Speaker 2: and Instagram. And there you really saw at scale all
[2:41] Speaker 2: the mistakes that meta was making all over the world.
[2:43] Speaker 2: And it's impossible not to pay attention to that stuff
[2:46] Speaker 2: once you've seen it.
[2:46] Speaker 1: Did Mark Zuckerberg and the Facebook execs you worked with
[2:50] Speaker 1: have any regard for the power they held and some
[2:53] Speaker 1: of the adverse effects that started to emerge with things
[2:56] Speaker 1: like addiction and the disinformation that we've seen explode?
[3:00] Speaker 2: You know, often these were problems that people paid attention to.
[3:03] Speaker 2: But then there was just a woeful lack of response
[3:09] Speaker 2: and a lack of investment in truly treating these things
[3:12] Speaker 2: as the priorities they should have been. Perfect example, right?
[3:16] Speaker 2: How is it that Facebook 20 plus years later doesn't
[3:19] Speaker 2: have any effective customer support team or function? Every day
[3:25] Speaker 2: there are millions of people on Facebook, businesses all over
[3:28] Speaker 2: the world, communities who have problems with Facebook, and they
[3:32] Speaker 2: find it impossible to get in touch with a real human.
[3:34] Speaker 2: at this company, which is one of the world's wealthiest,
[3:37] Speaker 2: you know, it's absolutely staggering. So for me, even if
[3:39] Speaker 2: you have leaders who say, yeah, this is a real problem,
[3:42] Speaker 2: we should solve it, clearly the commitments didn't live up
[3:45] Speaker 2: to the reality.
[3:50] Speaker 1: Let's move on to AI, because you joined Google DeepMind
[3:53] Speaker 1: in 2023, just as ChatGPT had exploded and generative AI
[3:58] Speaker 1: suddenly became something everyone was talking about. You said you
[4:02] Speaker 1: were brought back to Google to fix Google's storytelling on AI.
[4:06] Speaker 1: What was the story Google wanted to tell people about
[4:08] Speaker 1: what was coming?
[4:10] Speaker 2: I think all the big tech labs love to tell
[4:12] Speaker 2: the story. They love telling it up to the present day,
[4:14] Speaker 2: which is we're inventing this amazing technology which will allow
[4:18] Speaker 2: us to solve basically all of our problems. For decades,
[4:22] Speaker 2: mathematicians have been working to solve this problem. They haven't
[4:26] Speaker 2: been able to crack it.
[4:27] Speaker 3: But AI, it just did it easily.
[4:30] Speaker 4: OpenAI sending this out saying we're sharing a solution.
[4:33] Speaker 2: It doesn't matter what they are, you know, pick the
[4:36] Speaker 2: most heroic ones because, of course, you know, that's the
[4:39] Speaker 2: best way to tell a story often. A lot of
[4:41] Speaker 2: the focus would be on things like science and on
[4:44] Speaker 2: health care, on medicines. because those are things that are
[4:46] Speaker 2: really important for people's lives.
[4:48] Speaker 5: Melbourne researchers have developed a tool to detect the risk
[4:51] Speaker 5: of breast cancer using an AI algorithm. The breakthrough program
[4:55] Speaker 5: can provide.
[4:57] Speaker 2: One out of two of us is going to get
[4:58] Speaker 2: cancer in our lives. Everybody wants to, you know, live
[5:01] Speaker 2: in good health for as long as possible. We want
[5:03] Speaker 2: our children to have healthy lives. If you could use
[5:05] Speaker 2: AI to discover new drugs that cure cancers, that would
[5:09] Speaker 2: be a very good thing for the future, right? So
[5:12] Speaker 2: I think in their minds, that was a story that
[5:14] Speaker 2: was sufficient to get people excited about the trajectory of
[5:17] Speaker 2: this technology and all the things they were building. Of course,
[5:20] Speaker 2: it left out all sorts of other massive questions about
[5:24] Speaker 2: what does a future shaped by the most powerful technology
[5:26] Speaker 2: in history really look like?
[5:33] Speaker 1: That was the story they were telling us. What were
[5:35] Speaker 1: they telling each other inside these companies about generative AI?
[5:39] Speaker 2: I mean, in all the companies... The way that the
[5:42] Speaker 2: last few years has been presented has been as a race.
[5:45] Speaker 2: It is a race to achieve more and more advanced
[5:48] Speaker 2: AI against the other companies. So it's OpenAI, Google, Anthropic,
[5:53] Speaker 2: and so on. But it's also about nations.
[5:57] Speaker 1: We're leading China in AI.
[5:58] Speaker 5: We're the most sophisticated country in the world.
[6:01] Speaker 4: And frankly, I want to keep it that way because
[6:03] Speaker 4: whoever wins AI wins.
[6:06] Speaker 2: It's about the ecosystems of the United States. versus China
[6:10] Speaker 2: competing to be the leaders in global AI without thinking,
[6:14] Speaker 2: what is this truly in the service of? Is this
[6:16] Speaker 2: really going to serve the interests of most people, not
[6:18] Speaker 2: just the few people in these labs and in the
[6:21] Speaker 2: tech industry?
[6:22] Speaker 1: You said the way that we talk about governing AI
[6:24] Speaker 1: is fundamentally broken. So what are governments getting wrong?
[6:28] Speaker 2: Governments working with big tech have scoped most of the
[6:32] Speaker 2: problems of AI over the last few years as mostly technical.
[6:36] Speaker 2: with mostly technical solutions. So for example, this week, there's
[6:40] Speaker 2: a researcher at Anthropic who's been making global headlines.
[6:43] Speaker 4: Anthropic AI researcher, Jacob Coxon, who previously worked at OpenAI,
[6:48] Speaker 4: arguing that both companies are, quote, gambling with our lives.
[6:52] Speaker 4: But it's this post about his reason for stepping down
[6:54] Speaker 4: that's going viral, racking up more than 100 million views.
[6:59] Speaker 4: It reads in part, the people building AI earnestly believe
[7:02] Speaker 4: it could kill us all by the end of the
[7:05] Speaker 4: next decade.
[7:06] Speaker 2: This is nonsense. This is the kind of view that
[7:09] Speaker 2: comes from organizations where there's a very small number of
[7:13] Speaker 2: generally young men who are obsessed with science fiction trying
[7:16] Speaker 2: to peddle their fringe sci-fi thinking and turn that into
[7:19] Speaker 2: a full-scale agenda for how AI should be managed. Actually,
[7:26] Speaker 2: the real way we should be thinking about AI and
[7:29] Speaker 2: the true risks from AI is these are societal challenges
[7:32] Speaker 2: and societal choices that are at stake. There's no clever
[7:36] Speaker 2: algorithmic fix that's coming for the fact that AI will
[7:39] Speaker 2: end up displacing large numbers of jobs. It is not
[7:42] Speaker 2: a technical problem, the fact that AI is being used
[7:46] Speaker 2: now to turbocharge warfare.
[7:48] Speaker 3: Advances in technology are making it possible to develop autonomous
[7:52] Speaker 3: weapon systems that can pick and attack their targets without
[7:55] Speaker 3: human guidance. Around 200 autonomous weapon systems have been deployed
[8:00] Speaker 3: across Ukraine, the Middle East and Africa.
[8:02] Speaker 2: We are already seeing the huge effects, including in places
[8:05] Speaker 2: like Australia, of AI on consumption of resources and on
[8:09] Speaker 2: the environment, on our natural world.
[8:11] Speaker 6: This forecast from the Australian Energy Market Operator is saying
[8:15] Speaker 6: that in 10 years' time, these data centres, and there
[8:19] Speaker 6: are just dozens and dozens of them popping up around
[8:21] Speaker 6: the country, will consume 13% of the energy supplied into
[8:27] Speaker 6: the grid, the national electricity market, up from 3%, so
[8:30] Speaker 6: a massive jump.
[8:33] Speaker 2: So those are not technical problems. These are really choices
[8:37] Speaker 2: about how we choose to manage these things, about who's
[8:40] Speaker 2: paying the bills for these things, how we choose to
[8:42] Speaker 2: ensure that communities and their interests are respected. It's much
[8:47] Speaker 2: easier for the tech industry to say, these are all
[8:49] Speaker 2: technical problems. We can solve that if we just do
[8:51] Speaker 2: some unspecified things behind the scenes. Because surprise, surprise, that
[8:55] Speaker 2: puts power back in the hands of those very small
[8:57] Speaker 2: number of tech labs.
[9:01] Speaker 1: Still to come. Dex's warning for Australia. Dex, if we
[9:13] Speaker 1: look at the power balance here, you've worked very closely
[9:15] Speaker 1: with people running these companies. How much power do a
[9:18] Speaker 1: very small number of executives actually have over the direction
[9:22] Speaker 1: this technology takes? And do you think governments are in
[9:26] Speaker 1: a position to challenge them?
[9:27] Speaker 2: I think those small number of leaders have a huge
[9:30] Speaker 2: amount of influence over the technology. I mean, when you
[9:33] Speaker 2: look at some of the individual leaders, right, you can
[9:36] Speaker 2: see publicly the kind of capricious decision making by people
[9:39] Speaker 2: like Elon Musk being turned into product decisions for X
[9:44] Speaker 2: almost in real time. But, you know, when it comes
[9:46] Speaker 2: to the power that governments and regulators and communities have
[9:50] Speaker 2: over all of those companies, I actually think we have
[9:52] Speaker 2: a tremendous amount of power, much greater power than big tech.
[9:56] Speaker 2: We've simply chosen in many cases to exaggerate big tech's power.
[10:00] Speaker 2: or not to exercise the power that we do have.
[10:03] Speaker 2: Power is ultimately an illusion. It resides where people think
[10:06] Speaker 2: it resides. For a long time, a lot of political
[10:09] Speaker 2: leaders have tricked themselves into thinking big tech is untouchable,
[10:12] Speaker 2: and that's not true at all. They have to work
[10:14] Speaker 2: with regulators and decision makers all over the world if
[10:17] Speaker 2: they want to access our markets, if they want to
[10:19] Speaker 2: comply with regulations so that they can serve users in
[10:22] Speaker 2: lots of communities, if they want to work with advertisers
[10:25] Speaker 2: and partners. Ultimately, we've just got to be tougher in
[10:28] Speaker 2: asking for the things that we want to see reflected.
[10:30] Speaker 1: You say society is at risk of making the same
[10:32] Speaker 1: mistakes with AI as it did in the social media era.
[10:37] Speaker 1: What do you mean by that, Dex?
[10:39] Speaker 2: Well, here's the question. We're now more than a decade
[10:42] Speaker 2: and a half into the social media boom. I think
[10:45] Speaker 2: we've had just more time to learn how is social
[10:47] Speaker 2: media working and how is it not working, we have
[10:50] Speaker 2: spent years arguing over social media before we then very
[10:54] Speaker 2: quickly brought in these age restrictions for young people on
[10:57] Speaker 2: social media. Things which, in many cases, have yet to
[11:00] Speaker 2: be enforced well, so they are imperfect regulations. Nobody, or
[11:05] Speaker 2: very few people certainly that I'm aware of, have made
[11:07] Speaker 2: the case yet for, should we be exploring having age
[11:11] Speaker 2: restrictions on generative AI for young people? if you concede
[11:16] Speaker 2: that AI is going to be the most powerful technology ever,
[11:19] Speaker 2: this could have huge effects on the way young people think,
[11:22] Speaker 2: the way their brains develop, the way they process information,
[11:24] Speaker 2: the way they learn. Should we be just letting a
[11:27] Speaker 2: whole generation be guinea pigs for these tools, or should
[11:30] Speaker 2: we proactively bring in those restrictions now and then try
[11:33] Speaker 2: to understand what the effects are before we simply open
[11:35] Speaker 2: the floodgates?
[11:37] Speaker 1: Australia is actively courting the same AI companies you used
[11:40] Speaker 1: to work for. What should our government be demanding from
[11:44] Speaker 1: them before that technology becomes deeply embedded here, particularly given
[11:48] Speaker 1: we're talking about a handful of foreign companies that may
[11:52] Speaker 1: actually impact our national sovereignty?
[11:56] Speaker 2: I think government should be absolutely demanding transparency in how
[12:00] Speaker 2: their systems are operating, the things that will impact our lives, communities,
[12:05] Speaker 2: the interests of our communities in which this tech will
[12:09] Speaker 2: be operating. I think absolutely everything Governments should be also
[12:13] Speaker 2: shopping around. It is a true buyer's market. It is
[12:16] Speaker 2: an era in which the technology is being commoditized very,
[12:18] Speaker 2: very rapidly. When governments get obsessed with trying to cut
[12:22] Speaker 2: deals with just a very small number of hyperscalers, going
[12:25] Speaker 2: for these multi-billion dollar commitments in spending. These are things
[12:30] Speaker 2: where I think there's actually a massive global ecosystem out there.
[12:33] Speaker 2: There's plenty of other companies, there's plenty of great Australian
[12:36] Speaker 2: startups and companies who are building great bits of technology,
[12:39] Speaker 2: which in many cases are free from the kind of
[12:41] Speaker 2: baggage that comes from US big tech. So these are
[12:45] Speaker 2: all some of the big strategic questions. There's obviously a
[12:48] Speaker 2: whole lot of very practical questions around things like data sovereignty, privacy, security, IP, infrastructure.
[12:56] Speaker 2: But certainly, I think in general, decision makers should be
[12:58] Speaker 2: pretty tough in saying that we need you to abide
[13:01] Speaker 2: by certain things if you want to serve us. And
[13:03] Speaker 2: if you can't reach agreement on those, it's okay. You
[13:06] Speaker 2: can go your own way and we'll do business with
[13:07] Speaker 2: somebody else who respects our values and our interests.
[13:11] Speaker 1: And finally, Dex, you spent years helping these companies tell
[13:14] Speaker 1: us what the future could look like. You've said too
[13:17] Speaker 1: few people have a say in helping shape that future.
[13:20] Speaker 1: So how do we give workers and ordinary people a
[13:23] Speaker 1: real say in how AI changes their jobs and their lives?
[13:27] Speaker 2: I would like to see every decision maker, every leader
[13:33] Speaker 2: have a genuine open conversation now, put their cards on
[13:37] Speaker 2: the table about what they think is going to happen
[13:39] Speaker 2: with AI. AI is still largely a very elite conversation.
[13:43] Speaker 2: It's not a thing I think most people are talking about.
[13:46] Speaker 2: People will sort of hear the headlines and think, oh,
[13:48] Speaker 2: there's some stuff in the news. Every big tech leader
[13:52] Speaker 2: premises the vision of this AI future on the idea
[13:55] Speaker 2: that AI will make us all much wealthier. The world
[13:58] Speaker 2: will become more prosperous, economies will grow, there'll be more jobs.
[14:01] Speaker 2: Lots of people don't believe it. I'm very skeptical that's
[14:04] Speaker 2: how it will unfold in the current way it's being organized.
[14:06] Speaker 2: But certainly, if your vision is that this will help
[14:09] Speaker 2: grow Australia's economy, then as a country, as communities, we
[14:13] Speaker 2: should get a question asked. What do we want to
[14:16] Speaker 2: do with all that wealth? How should we be using it?
[14:19] Speaker 2: What should we be allocating it for? How should it
[14:21] Speaker 2: be distributed? And so fundamentally, there are massive choices now
[14:25] Speaker 2: about our futures, about our children's futures. And we've got
[14:29] Speaker 2: very little time in which we might get to make
[14:30] Speaker 2: some very big choices about what kind of country we
[14:33] Speaker 2: want to live in, what kind of people do we
[14:35] Speaker 2: want to become in the future. And so this is
[14:38] Speaker 2: genuinely a conversation. We've just got to have it. Somebody
[14:40] Speaker 2: has to lead it. Um, and I think, um, anyone
[14:43] Speaker 2: who can help shape that and open it up, I
[14:45] Speaker 2: think has a responsibility to do that.
[14:47] Speaker 1: How much time do we have to have that conversation?
[14:50] Speaker 2: We have very little time for it because of course, uh, every,
[14:54] Speaker 2: you know, major AI lab will tell you, and I
[14:56] Speaker 2: do agree with them on this, that we will arrive
[14:59] Speaker 2: at a much more advanced form of AI, artificial general intelligence,
[15:03] Speaker 2: essentially human level intelligence and machines sometime in the next
[15:07] Speaker 2: 10 years. Most of those leaders believe it could be
[15:11] Speaker 2: closer to five years from now. OpenAI's leadership has just
[15:15] Speaker 2: been signaling in the last few days they believe they
[15:17] Speaker 2: may have already achieved it. Lots of disagreement about that.
[15:21] Speaker 2: Many people, including myself, think that's just marketing hype. And yet,
[15:24] Speaker 2: when that arrives, if we have not fundamentally reassessed some
[15:28] Speaker 2: of our thinking about our economies, about our societies, about
[15:32] Speaker 2: our institutions, our governance, then we will be out of
[15:35] Speaker 2: time and the technology will end up deciding for us.
[15:38] Speaker 2: Which is exactly what happened with social media when we
[15:41] Speaker 2: sleepwalked through a whole generation of technology, but we hadn't
[15:44] Speaker 2: taken it seriously before we entered that new era. And
[15:47] Speaker 2: I just don't want us to make that same mistake again.
[15:52] Speaker 1: Well, Dex, we're out of time for this conversation, but
[15:54] Speaker 1: thank you so much for yours.
[15:55] Speaker 2: Thanks for having me.
[15:58] Speaker 1: Dex Hunter-Torrick has been in Australia for the Philanthropy Australia Conference.
[16:14] Speaker 1: Also in the news, Pauline Hanson has offered a half
[16:17] Speaker 1: apology for her comments describing Indigenous Australians as primitive. saying
[16:22] Speaker 1: she's sorry if she's offended people.
[16:25] Speaker 5: We need to be united as a nation and stop
[16:27] Speaker 5: the division that's happening by politicians.
[16:29] Speaker 1: But the argument is that you're worsening it by saying
[16:31] Speaker 1: things like that they were the most primitive race on
[16:34] Speaker 1: earth at the time.
[16:35] Speaker 6: That that upsets a lot of Indigenous Australians.
[16:40] Speaker 5: I don't... Well, then I regret saying that because I
[16:43] Speaker 5: don't want to upset people. That's not my intentions. That
[16:48] Speaker 5: wasn't my intentions to hurt people.
[16:51] Speaker 1: Hanson also offered a half apology to the Prime Minister,
[16:54] Speaker 1: if he's taken offence, to her comments about his mum
[16:57] Speaker 1: living in a housing commission. Gold Coast Mayor Tom Tate
[17:02] Speaker 1: has sworn at a journalist and stormed out of a
[17:04] Speaker 1: press conference and made questions over a dinner he claims
[17:07] Speaker 1: to have had with Donald Trump earlier this year.
[17:10] Speaker 5: Why are you so aggressive?
[17:11] Speaker 1: Are you from the ABC?
[17:11] Speaker 2: Yes.
[17:11] Speaker 6: Yeah, f * * * off.
[17:12] Speaker 1: Excuse me?
[17:12] Speaker 2: F * * * off. I'm serious. I'm serious.
[17:18] Speaker 1: A statement from the Trump organisation claims the mayor's account
[17:21] Speaker 1: of what occurred is a complete fabrication. I'm Daniel James.
[17:25] Speaker 1: Thanks for listening to 7am. We'll be back tomorrow.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.