It has been a busy few weeks in the world of AI, with apocalyptic predictions back front and center after the latest “whistleblower” former Anthropic researcher Jacob Coxon quit publicly, declaring that tech companies were “racing straight to superintelligence and gambling with our lives.” His post racked up 153 million views in thirty-six hours. Right on cue, AI CEOs clamored in, apparently with a renewed enthusiasm for the very regulation they have been literally skirting for years. Anthropic CEO Dario Amodei called for binding safety testing, while Google DeepMind’s Demis Hassabis now apparently wants an international body inspecting models before release.
The elephant in the room is that this is precisely what the EU AI Act which gives regulators authority to inspect, evaluate, and approve high-risk AI systems and foundation models before market release was intended to do. And you guessed it: their compliance has not exactly been consistent. While major AI developers (OpenAI, Anthropic, Google, and Meta) have signed voluntary agreements and participated in high-profile pilot tests, their actual submission of frontier models for pre-release inspection has been selective, inconsistent, and often fraught with tension. In other words, they are failing to submit to the very regulations they claim are desperately needed to save us all from the risk of a god-like singularity—one that promises either utopia or the end of the world apparently, and which lies on the other side of AGI, the so-called “superintelligence” they all fantasize is coming. Think, the rapture but for tech nerds.
The case for the apocalypse rests on two core ideas: first, machines pursuing goals their human creators never intended; and second, a system getting smart enough to rewrite its own code, accelerating far beyond our ability to control it. Both sound terrifying until you realise they are currently based on speculative risk, the “trust me, bro” Silicon Valley hype around what might happen, while ignoring the very real risks happening right now.
That is why, when pressed in an interview on CNN about his catastrophizing predictions, Jacob Coxon admitted: “Right now, no. The models we have today are not an existential threat to humanity.” He went on to clarify: “It’s not about today’s systems taking over the world tomorrow. Today’s risks are things like deepfakes, automated misinformation, and cyber vulnerability. The existential concern is about what happens two, three, or five generations down the line as these systems gain autonomous agency.”
He could have added “if” they develop an autonomous agency, because that, too, is entirely speculative. Where AI agents have been seen performing what is interpreted and often problematically framed in the media as ”autonomous” tasks, such as the 2023 story of an AI agent emailing Cambridge philosophy professor Dr. Henry Shevlin asking for work, it is simply executing an automated recursive loop: evaluating its own output against a prompt, deciding the next step, and executing code. It is not an indication of consciousness, though it raises questions around the risks of highly effective phishing emails and broader cybersecurity concerns where teams unleash technologies online whose inner workings they do not fully understand. Dangerous? Yes. Sentient? No.
Mainstream reporting frequently loses these critical details. Hysteria around killer robots, superintelligent beings, or immortal transhumans makes for far more compelling reading and clicks than sober risk assessments.
Underpinning so much of this spin is the Kool-Aid Silicon Valley has already drunk: that they are working on the new frontier of transhuman or post-human intelligence. Unbeknownst to many is the cult-like thinking dominating Silicon Valley, where many believe the technology will eventually deliver a god-like intelligence a tech rapture via the singularity. If you want to understand this pseudo-religious tech cult, I highly recommend astrophysicist and science journalist Adam Becker’s book, More Everything Forever: AI Overlords, Space Empires, and Silicon Valley’s Crusade to Control the Fate of Humanity. Becker investigates these quasi-religious ideologies namely Longtermism, Effective Altruism, Transhumanism, and AI Eschatology exposing how “billionaire prophets” cloak dangerous, self-serving fantasies of digital immortality, space colonization, and AI godhood in the language of science to justify power, distract from real-world issues, and prop up astronomical valuations.
This cult has global followers, including disgraced former Oxford fellow Nick Bostrom, who ran the now-defunct Future of Humanity Institute at the University of Oxford until emails from his graduate student years resurfaced referencing explicitly racist tropes and defenses of eugenics. Bostrom issued an apology, but it failed to placate critics largely because he conspicuously declined to retract his core premises regarding race and intelligence, choosing instead to offer a defensive academic pivot toward eugenics and population genetics. Yikes.
At the center of Bostrom’s worldview is the conviction that rapid advances in artificial intelligence will inevitably yield a “superintelligence” a belief rooted in hierarchies of human intelligence (of which tech elites, naturally, view themselves as the pinnacle) and the emergence of an intellect that vastly surpasses human thought. And wait for it: an intellect that could also be sentient, which explains the surge of legitimate-sounding articles claiming a language model can “feel pain,” love, or experience emotion. It cannot. It is code.
A committed transhumanist, Bostrom frames superintelligence not as science fiction, but as a crusade for human enhancement championing cognitive modification and even age reversal through unchecked technological application. It is precisely this high-tech messianism that allowed Bostrom’s ideas to capture Silicon Valley and earn praise from figures like Bill Gates, Elon Musk, and Peter Singer, effectively disguising deeply troubling, elitist assumptions about human value under the shiny veneer of technological progress a progress that can neither be controlled nor valued by normal financial metrics, because it is running on supremacist cult vibes disguised as smart nerd talk.
Cut to the AI bubble, which was dominating headlines before this latest round of “AI is going to kill us all” frenzy was relaunched diverting attention away from overhyped valuations and the fragile economy toward kicked-down-the-road IPOs and “arms race” rhetoric. Who can forget former Governor and current United States ambassador to Israel since 2025, Mike Huckabee discussing AI and national security during a studio interview with The Media Line in Jerusalem on September 17, 2026, in which he spells out the line propping up the current artifice: “It is an arms race, pure and simple. If America decides to slow down or hit the brakes on AI because of fear or over-regulation, China will hit the accelerator. We don’t have the luxury of opting out.”
While the race to produce powerful AI acts as cover for defense companies seeking to cash in on new technologies, the picture out of China does not support the idea that Beijing buys into the existential threat narrative so frequently propounded in Western media. When you look at China’s actual AI policy, Beijing’s stated fears center overwhelmingly on deepfakes, labor displacement, and political destabilization not a sci-fi superintelligence extinction event. If AGI were the existential threat tech leaders claim, wouldn’t every nation developing AI cite it as a primary concern?
So much of the AGI risk narrative rests on assumptions and hocus-pocus predictions. In reality, we have advanced predictive models ”stochastic parrots,” to quote the metaphor introduced by Dr. Timnit Gebru and her co-authors to describe how large language models (LLMs) stitch together words based on probability without comprehension. The term originated in a landmark 2021 research paper co-authored by Gebru, Emily M. Bender, Angelina McMillan-Major, and Margaret Mitchell, which warned of LLM dangers that look nothing like today’s sci-fi hype. At the time, Dr. Gebru was the most senior Black woman on Google’s Ethical AI team when the company forced her out after rejecting the paper which has since become one of the most cited works in AI research.
Their paper warned against massive environmental costs, uncurated training data, and the amplification of societal biases and misinformation in large models, demonstrating how these tools reinforce existing racial and gender inequalities. It is striking that while the harm of AI is already non-theoretical for millions of people across the globe, the “risks” dominating public debate remain entirely speculative. Meanwhile, immediate dangers are ignored. AI was involved in military targeting in the genocide in Gaza and the bombing of the girls’ school in Iran (which our apparently super clever AI deduced was a military base!). UN reports show AI data centers consuming more electricity than entire nations like Saudi Arabia, while studies reveal our growing reliance on these tools is actively degrading human cognition. Where is the hysteria over these issues?
Worse still, where is the scrutiny of these over-inflated companies and their incestous ties. Determining whether an AI boom or bust is underway is made harder by the web of circular financing between companies that act simultaneously as investors, suppliers, and customers for one another. Microsoft is both OpenAI’s primary investor and its chief commercial partner. In its 2026 tax filings, Microsoft reported $24.1 billion in revenue tied to commercial arrangements with OpenAI—accounting that AI skeptic and bestselling author Cory Doctorow describes as “really weird math.” The five largest U.S. tech giants are on track to pour over $700 billion into AI infrastructure this year alone yet independent analyses indicate the industry faces a $600-billion gap between capital expenditure and software revenue. Critics like hedge fund manager Michael Burry call the doomsday narrative “self-serving hype and puffery” a convenient storyline to prop up astronomical valuations as Anthropic prepares for a $965-billion listing and OpenAI pushes its IPO down the road. Many economists predict they are due a serious devaluation which could impact the entire, already flailing economy.
On The Tea, we love to bring you grounded, solid voices in this space. We recently heard insights from Cory Doctorow—bestselling author and creator of the concept of “Enshittification,” whose latest work includes A Reverse Centaur’s Guide to AI. Cory remains deeply skeptical of these exaggerated claims so I highly recommend his recent episode where he debunks much of the marketing hype. You can also check out last week’s show with former Irish diplomat and Oxford professor Dr Jennifer Cassidy, warning about much more tangible threats to our national security posed by the handing over of our data to the silicon valley bros working on experimental tech our own governments don’t have a handle on, in what she dubs ‘an open scandal’ in diplomatic circles - much tea spilt on digital sovereinty. You can also check out our episode with Oxford professor of Philosophy, Carissa Véliz who argues that AI prediction and mass surveillance undermine human agency, democracy, and privacy by turning algorithms into instruments of behavioral control and social manipulation or with child prodigy and founder of STEMETTES Dr Anne-Marie Imafidon MBE argues that our growing over-reliance on generative AI tools is quietly altering human cognitive development, eroding foundational critical thinking skills, and outsourcing fundamental social intelligence from a very young age. Finally, if you want to hear an alternative perspective, check out one of our most watched videos with the co-founder of Stability AI Emad Mostaque on why he thinks AI could end capitalism (and possibly also the world!). All the links are on our YT - let us know where you land!
You can never take these AI promotors seriously. All of a sudden they are coming up with some narrative. AI is dangerous today in many ways. It will never kill humanity. But it has to be regulated.