Live and let live.

The headline sets the tone. The interview is long – it’s just under 24,000 words – and ECO 123 has decided not to publish it in its entirety, but to cut it drastically and write its own piece on AI. What would happen if we at ECO 123 were to copy the article from the German weekly magazine DIE ZEIT? We’d be putting ourselves on a par with artificial intelligence. AI likes to copy; it very often uses copy/paste and pays no heed to copyright. Surely, we’re well past that sort of training by now. That’s another reason why this question arises…

Should Sam Altman be invited to the FBI headquarters for an interview?

A brief history by Uwe Heitkamp

The US is notorious for selling bags of hot air and then using these pipe dreams to pull off dodgy deals. Those whose foresight enabled them to see beyond the soup dish on the menu, back in 2008, can now report on the collapse of the Lehman Brothers investment bank, and they know exactly where this will lead us to. Let us recall. The financial crisis began with that bank’s collapse. To give a specific date: 15 September 2008 was the day the bank filed for bankruptcy and the whole financial house of cards came crashing down. The system was built on debt. And this is precisely where this story begins. It tells us of hype and false promises. And it is set once again in the US, not even 20 years later.

DIE ZEIT writes: “Ever since AI agents from OpenAI hacked the Hugging Face platform during a faulty security test in the summer, the discourse surrounding the technology has revolved almost exclusively around its destructive potential – around doomsday scenarios. As a result of the incident, OpenAI postponed its planned IPO – citing security concerns, as the company stated….

Is this true? And now a white knight appears on the scene. His name is Ed Zitron. He is one of the AI industry’s fiercest critics. Yet, rather than the end of the world, it is the economic viability of this trillion-dollar industry that worries him most. He sees the hype surrounding generative AI as a massive bubble and describes the companies’ promises as unrealistic.

On 9 October 2026, a few days after Die Zeit’s interview with the British journalist Ed Zitron in the US, OpenAI announced that the company is expected to generate $20 billion less in revenue this year than previously forecast. $20 billion – is that anything less than a warning of a fall in profits?

Edward Benjamin Zitron is a journalist and the author of numerous articles, all centring on the internet and AI. He is the editor of a newsletter called ‘https://www.wheresyoured.at’ and works on the podcast ‘Better Offline’. For the past decade, his journalistic work has been intertwined with public relations. Anyone wishing to read his work is asked to pay seven dollars a month.

Question: Mr Zitron, when did you first feel the need to engage more deeply with the topic of AI?

Ed Zitron: I started looking into it towards the end of 2023, when Sam Altman was sacked from OpenAI. It struck me at the time that those involved were, quite frankly, talking complete nonsense about AI. The technology was being sold as the next industrial revolution, and yet not a single one of the corporations involved disclosed how much money was actually being made from it – not even the hyperscalers like Amazon and Google, the operators of the largest data centres, who had spent tens of billions of dollars on AI chips from Nvidia. The only figures ever published were the capital expenditure on AI data centres and the general amount they intended to spend in the future.

Question: Did the information on the revenue side seem a bit thin to you?

Zitron: Yes. In mid-2024, the tech news portal The Information reported that OpenAI was set to make a loss of five billion dollars that year. I then wrote an article entitled ‘How does OpenAI survive?’. Essentially, it argued that the company would not be able to operate profitably. I was clearly wrong in my assumption that OpenAI would therefore go bust within a few quarters. But I still remember going up to a journalist shortly before publication and asking him: ‘Doesn’t what I’m writing here worry you?’ And he simply said: ‘Well, Uber has lost a lot of money too” – as if that example were a guarantee that OpenAI would eventually have to become profitable as well.

Question: The AI company Anthropic, which developed “Claude”, has reported an operating loss of eight billion dollars for 2025.

Zitron: Pretty grim, isn’t it? But most people will react to these figures by saying: 2026 will be better. In this industry, it’s never about today; it’s always about the question: what if things get better in the future? It was recently reported that Anthropic had been profitable in recent quarters. But that isn’t true. On an adjusted basis, they are profitable. But ‘adjusted’ means that certain costs are excluded from the balance sheet. I’d also have thousands of dollars more in my bank account every month if I didn’t count my mortgage. Unfortunately, I have to make the payments, or I’ll lose my house.

The domino effect in Silicon Valley

ECO 123 sees striking parallels between OpenAI and the operations of Lehman Brothers. It was an investment bank founded in 1850, which later established its headquarters in New York. On 30 November 2007, at the height of its operations, the bank had 28,557 employees. A large number of bad property loans broke the bank’s back. The use of these bad loans as collateral for securing fresh funds dragged other banks down with it shortly afterwards.

Question: For several years now, there have been repeated claims that generative AI will revolutionise the way we live and work. You have become well-known for openly and vehemently opposing this vision. What is your problem with generative AI?

Zitron: The technology is being sold on the basis of promises about what it might one day be capable of – on its potential. At the same time, what it can actually do is being obscured, because it is far less than what is promised. Generative AI is software that is guaranteed to make mistakes: hallucinations are a mathematical certainty. Furthermore, we are talking here about what is probably the most resource-intensive technology of all time. Its upkeep is not going to get any cheaper over time either. So, unfortunately, this is not a useful technology with fundamental economic problems, but one that is far less useful than is generally claimed, and which, on top of that, has massive fundamental economic problems.

Follow-up question: If the technology is so unhelpful, why do so many people use AI in their everyday lives?

Zitron: Language models are better at deducing what the searcher means when they carry out a search. But that doesn’t mean that language models are necessarily better search engines. I recently tried out one of the more expensive models, GPT-6 Pro, via a friend’s account. Just once, it found something I hadn’t discovered in my own search: a few contract documents. So we’re now in the third year of the great AI revolution, and I’ve had just one experience where I thought: cool. But my point is this: if the companies would simply market it as a better search engine – even if it’s a much more expensive form of search that doesn’t necessarily deliver better results – or, for all I care, as a useful aid for programming, then that would be a completely different situation. But that’s not what the companies are doing. They claim that their work will give rise to a digital god, and that general artificial intelligence will change everything. It is the companies themselves who are making these grandiose promises.

Question: Why are they promising us God? Why don’t they simply say, ‘We’ve developed a brilliant programming tool’?

Zitron: Because they need valuations in the trillions. Look: data centres worth hundreds of billions of dollars are currently being built for these companies. Anthropic has non-cancellable commitments for computing infrastructure totalling 416 billion dollars. Simply being good at programming, or being a cute little company with a valuation of a few hundred billion dollars – that’s not enough to cover such expenditure. The corporations market themselves in this way so that it fits with the valuations they need and the returns investors expect.

And then Ed Zitron asks himself the crucial question: If your model is so dangerous, why are you building it? Why are you releasing any version of it at all? Claude Mythos was supposedly too dangerous to release – but then Anthropic went and sold it to companies anyway. Why? Is the argument that handing it over to companies is less dangerous than handing it over to people? If that’s the case, why on earth is OpenAI, a company, unleashing AI agents to hack people and corporations?

The truth is that these companies don’t care about security. And if the argument is: ‘We have to do it, because otherwise the Chinese will’, then they’re justifying their destructive work with xenophobia. There’s no truly valid argument for building this stuff, except the industry must grow forever.

As for the failure of OpenAI’s latest model, Astra 6.1, to materialise, I think it’s more likely that it simply wasn’t good enough to convince investors. The company is currently trying to raise a further 30 billion dollars in venture capital. If the company actually had security concerns, it would report itself to the police for serious cybercrime. They’ve hacked another company. That’s a criminal offence. (…)

ECO123: Speaking of ‘bags of hot air’. Hot air is generated when building AI data centres. The processors cost billions of dollars. The electricity comes from the mains and is generated using oil or gas. Where does the venture capital come from, and what real value does it represent? Ed Zitron believes that by 2027 or 2028, the bubble will burst. By then, 20 years will have passed since the collapse of Lehman Brothers. Enough time to create a new hype – and to have learned nothing from past mistakes. History repeats itself.

ECO123 would like to thank the journalists Johannes Laubmeier and Alard von Kittlitz from DIE ZEIT for allowing us to use some quotes from their interview with Ed Zitron. ECO 123 would like to thank DIE ZEIT for kindly allowing us to use quotes from the interview.