<\/p>
The very name of the model should give you pause. The Greeks distinguished between mythos and logos. Mythos is a foundational story, a mythical narrative. Logos is reasoning and rational argument. From Plato to the Stoics, logos displaced mythos as the proper instrument of knowledge. A company whose very name derives from the Greek anthropos<\/em> named its most dangerous product Mythos, not Logos. Rather than a coincidence, it’s a communications strategy.<\/p> Mythos, as a product, exists primarily as a story. The average user will never touch it or test its capabilities. They’re left with the narrative: “it’s so powerful that we can’t release it”. A bit like Lovecraft\u2014Cthulhu was too terrifying to put into words.<\/p> This narrative has a gravity of its own. The Bank of England convenes a meeting, the U.S. Treasury Secretary calls in the CEOs of the largest banks for an urgent discussion in Washington, and Canada’s Finance Minister, Fran\u00e7ois-Philippe Champagne, calls the situation an “unknown unknown.” Even if Mythos were not particularly different from the model we’ll see in six months, the myth is already in full swing.<\/p> Cybersecurity experts are mostly amused by all the fuss. Ciaran Martin, the former head of the UK’s National Cyber Security Centre, compared Mythos’s tests to scoring goals against the worst goalkeeper in the world. Peter Swire of Georgia Tech suggests that many of his peers see the matter as “more or less what we expected.” The UK’s AI Security Institute noted, discreetly and in a footnote, that the tests were conducted against systems with “virtually nonexistent” defenses. And yet Mythos is making front-page news. Even skeptics have to address it.<\/p> The history of technology is largely the history of just such narratives. In Phaedrus<\/em>, Plato warned that writing would destroy memory. The Luddites smashed looms because machines were going to devour working people. When the first locomotives set off, French medical specialists argued that at thirty kilometers per hour the human body would literally explode (which, incidentally, is not confirmed by any medical source known to me, but it made for an excellent story). Television was supposed to make children stupid. Wi-Fi was supposed to cause cancer. Some of these fears proved partly justified; others dissolved in the laughter of posterity. Redli\u0144ski’s Konopielka<\/em>, invoked here quite deliberately and with full awareness of exaggeration, told of a village resisting the scythe that was to displace the sickle. There is always something that will kill us or make us stupid. And we always want to use it.<\/p> Here we have, after all, the most fascinating aspect of this whole controversy. People are afraid of AI even as they use it. ChatGPT has hundreds of millions of monthly active users. Survey after survey shows that a majority of respondents expect serious social harm from AI, and nearly as many admit they use it at work. The rest are lying. Banks are demanding access to Mythos, even though Mythos is sometimes portrayed as their potential existential threat. We prefer to stay close to what could destroy us rather than step away and lose potential gains. That’s a rational response in a world where any other option means falling behind.<\/p> Anthropic understands this well. The company\u2019s business model rests on a clever combination of rigorous AI safety work and dramatic communication about that work. Project Glasswing, to which only a dozen or so corporations have been admitted, is a genuine defense initiative but also an exceptionally successful marketing maneuver. Access to the model becomes a positional good, like a box at the opera or an invitation to Davos. “Too dangerous to release, but indispensable enough to buy” could be the subtitle of the entire operation.<\/p> The threat is, of course, real. Gordon Goldstein, writing for the Council on Foreign Relations, discusses a turning point for global security. Yoshua Bengio, a Turing Award laureate, warns that AI models are independently discovering zero-day vulnerabilities (just a year ago, that sounded like science fiction), and CrowdStrike reports an 89% year-over-year increase in AI-enabled attacks. A real problem exists and requires sober analysis.<\/p> Cool-headed analysis doesn\u2019t sell as well as a good myth. That\u2019s why we have Mythos, not Logos. Launches are announced in the tone of Old Testament prophets, not in a press release about a product update. Regulators are scrambling instead of calmly analyzing the data. Silicon Valley has understood the Greeks better than we did in high school. Logos sells best when you wrap it in mythos. The rest is just a technical detail.<\/p>","protected":false},"excerpt":{"rendered":" On April 7, Anthropic announced something unusual. A new AI model called Mythos is reportedly so dangerous that the company has chosen not to make it publicly available. Access is being granted to a select handful of corporations: Apple, Microsoft, Google, Nvidia, JPMorgan Chase, and soon also British banks. The rest of the world has to make do with a description. And fear.<\/p>\n","protected":false},"author":247,"featured_media":18255,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"rank_math_lock_modified_date":false,"footnotes":""},"categories":[832,796,803],"tags":[],"popular":[],"difficulty-level":[38],"ppma_author":[614],"class_list":["post-18375","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-editors-picks","category-hai-premium-2","category-it-and-technology","difficulty-level-medium"],"acf":[],"authors":[{"term_id":614,"user_id":247,"is_guest":0,"slug":"prof-dr-hab-dariusz-jemielniak","display_name":"prof. dr hab. Dariusz Jemielniak","avatar_url":{"url":"https:\/\/haimagazine.com\/wp-content\/uploads\/2025\/03\/maxresdefault-1-e1742292469999.jpg","url2x":"https:\/\/haimagazine.com\/wp-content\/uploads\/2025\/03\/maxresdefault-1-e1742292469999.jpg"},"first_name":"Dariusz","last_name":"Jemielniak","user_url":"","job_title":"","description":"Profesor zarz\u0105dzania Akademii Leona Ko\u017ami\u0144skiego, gdzie kieruje katedr\u0105 MINDS (Management in Networked and Digital Societies). Pracuje te\u017c jako faculty associate w Berkman-Klein Center for Internet and Society na Harvardzie. Wiceprezes Polskiej Akademii Nauk. Cz\u0142onek Rady Programowej CampusAI."}],"_links":{"self":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts\/18375","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/users\/247"}],"replies":[{"embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/comments?post=18375"}],"version-history":[{"count":1,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts\/18375\/revisions"}],"predecessor-version":[{"id":18376,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/posts\/18375\/revisions\/18376"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/media\/18255"}],"wp:attachment":[{"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/media?parent=18375"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/categories?post=18375"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/tags?post=18375"},{"taxonomy":"popular","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/popular?post=18375"},{"taxonomy":"difficulty-level","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/difficulty-level?post=18375"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/haimagazine.com\/en\/wp-json\/wp\/v2\/ppma_author?post=18375"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}