Posted: Thu Oct 08 2026 | Last Updated: Thu Oct 08 2026
Don’t believe the criti-hype. Jacob Coxon, a 28-year-old researcher at Anthropic, [recently quit the company](https://www.bbc.co.uk/news/articles/c1kx0gyje9wo) due to fears that artificial intelligence (AI) will wipe out humanity. How might that happen, exactly? Coxon provided little by way of detail. Evidence to back up the claim? None.
Evan Hubinger, alignment science lead at Anthropic, echoed the claims of his former colleague and said that, actually, yes, there’s a [10% chance](https://www.bbc.co.uk/news/articles/ckgwy1k42w4o) AI systems will ‘kill all humans’ in the next, say, ten years. How might that happen? Recursive self-improvement (RSI). Evidence of RSI? There is none. A blockchain of Tech Bros soon jumped on the apocalyptic bandwagon, suggesting 10%, 20%, maybe even 30% chance of extinction.
We all love a nice round number. But perhaps the most interesting part of the criti-hype prediction cycle is the reaction, or lack of effective reaction, from the media. Journalists parrot the round numbers and seldom ask for evidence. Some journos stood aghast, as they interviewed Coxon. Others did that Owen Wilson ‘wow’ thing. But very few effectively scrutinised the claims.
It’s particularly galling because criti-hype, the use of hyperbolic self-criticism to promote the power of tech, has a long and dull history. Tech Bros have previous: raising various alarms to showcase the supposed potency of AI. They told us years ago that AI will take our jobs, that we’re facing a blue-collar bloodbath. [Unimaginable risks were imminent](https://www.washingtonpost.com/business/2023/05/30/ai-poses-risk-extinction-industry-leaders-warn/ "Unimaginable risks were imminent"), five years ago, three years ago, last month, yesterday. Loss of control is a constant fear, say the people in control.
None of that has come to pass. Nor have the more absurd tech predictions, going back decades. Bill Gates said [Spam](https://www.theguardian.com/technology/2004/jan/25/billgates.spam) would be eradicated in two years. That was in 2004. Zuckerberg said we’d soon work, learn, shop, and play in the Metaverse but, to the best of my knowledge, I’m no avatar. A man will walk on mars in ten years, [Musk said](https://www.wsj.com/articles/BL-VCDB-10984), fifteen years ago. Now Sam Altman says AI will ‘fix’ the climate crisis (it’s making it worse) and lead to ‘Universal extreme wealth for everybody’ (lend me money). We’ve long had to suffer their compendium of nonsense.
Tech Bros use disinformation to promote tech that [promotes disinformation](https://pulltheplug.uk/the-risks-of-ai). It’s all rather cozy. George Orwell described journalism as the art of printing something that someone does not want printed. Everything else is PR. Companies would pay a high price to receive a tenth of the column inches that AI Tech Bros receive. They make ‘scary’ predictions – self-promotion masquerading as self-criticism – and journalists repeat it verbatim. The public has to suffer them.
Criti-hype, parrotted by the media, dilutes our judgment. Citizens, apparently on the verge of extinction, no longer know what to believe. We’re becoming increasing suspicious of inflated claims. Tech Bros claim existential risk and people suspect they’re masking [everyday harms](https://pulltheplug.uk/the-risks-of-ai), such as copyright theft, the introduction of bias, and environmental damage. A new ‘catastrophe’ arrives and I have no idea whether it’s actually catastrophic or whether an IPO lurks around the corner. They call for regulation – supported by [89% of the public](https://www.adalovelaceinstitute.org/policy-briefing/great-expectations/) – and we start to suspect [regulatory capture](https://www.wired.com/story/why-anthropic-is-pushing-states-to-regulate-ai-faster/): introducing legislation that raises costs to hider rivals.
It’s frustrating. It’s particularly frustrating because the existential threat is very real, even if not dystopian or apocalyptic. The risk of rogue AI building bioweapons or attacking critical infrastructure poses serious risks. But the calamitous predictions, the perpetual hyperbole, the constant use of round numbers dilutes the substance of those risks.
And it’s worth mentioning that the same people making the predictions, parrotted by the press, largely stand in the way of mitigating risks. They remain [secretive about datasets](https://www.theatlantic.com/technology/archive/2023/10/ai-technology-secrecy-transparency-index/675699/), shielding themselves from IP lawsuits. They do not – or [indeed cannot](https://www.nytimes.com/2026/04/15/magazine/ai-black-box-interpretability-research.html) – provide visibility about how their models work, even if visibility would mitigate the worst of the risks.
AI models exist in darkness. And that darkness is secured by the same people who tell us their tech may kill us all. It’s absurd. For a democracy to properly function, for [citizens to make informed decisions](https://pulltheplug.uk/citizens-assemblies), we need trustworthy information. We need a free press that scrutinises bad actors, that holds them to account. To effectively regulate AI, we need a democratic consensus, based on verifiable and trustworthy information.
Tech Bros have long been crying wolf and increasingly the public do not know what to believe. That does not mitigate the harms of AI. It extends them. Democracy demands visibility so that we can call for the right regulations and guardrails. At the moment, the public has to suffer childish hype men, boys crying wolf, an absurd eco-system of criti-hype and self-promotion masked as caution. It’s a disservice to us all. And it needs to end.
---
## [**Join Our March**](https://pulltheplug.uk/forms/sign-up-to-march-against-the-machines)