Anthropic – Real Context News (RCN)

The Real Context News AI Reading List: EMERGENCY Anthropic MELTDOWN Edition (#8)

Brian E. Frydenborg — Fri, 05 Jun 2026 23:18:05 +0000

“Skynet is here…ish? Maybe? Who knows! Time to PAUSE!”—Anthropic

(Traduce en español/translate to Spanish)

By Brian E. Frydenborg (Twitter @bfry1981, LinkedIn, Bluesky, Facebook, Substack with exclusive informal content) June 5, 2026; because of YOU, Real Context News surpassed one million content views on January 1, 2023, but I still need your help, please keep sharing my work and consider also donating as I make my overdue comeback! Real Context News produces commissioned content for clients upon request at its discretion. All other editions of the Real Context News AI Reading List can be found here. And be sure to check out my related book review of Nicole Perlroth’s must-read, award-winning book on cyberwarfare as well as my own work on cyberwarfare.

A bit early… Terminator 2: Judgment Day—James Cameron/Tristar Pictures/Lionsgate

Dear readers, I’m 44 years old. And I’ve never seen anything like this (could apply to multiple stories but hey, let’s do this one).

In the span of about not even two weeks, the following has just happened:

We had a pope come out with a moral and ethical wakeup call at a time when one wondered if the Catholic Church was ever going to truly matter in a profound way on the international stage again.

That in itself is truly a miracle.

Yes, Pope Leo XVI roared like his namesake predecessor and like the lion he is that, yes, there was still going to be profound moral leadership in a global sense in the era of Trump 2.0, and he did so with one of AI giant Anthropic’s co-founders, Chritopher Olah, who, despite looking like a teenager, is a relative moral titan for his industry in joining the pope in what I dare say is sounding like a crusade (maybe the first good one). Leo uses the word “disarm” for AI, and he made it damn clear he was not mincing words.

Then, about a week later, at the very end of a long update from The Anthropic Institute that very much reads like a research paper, I see something I’ve never really seen before from a company describing its own technology. They are flat-out telling use they have lost control of their experiment, and it may have become Skynet as in T2 Terminator 2 (just as I warned when I launched this List, more humorously than now, and not thinking it was possibly a few months away), that they basically don’t know if we can even know for sure, are clearly having a moral crisis of conscience at being one of the driving forces behind this possible calamity, and are desperately trying to put on the brakes and, at least for now, saying we need to stop to be safe or a sentient AI may very well deem us a threat and, well… watch the movies. They—some of them barely not kids—sound like they’ve lost control of their version of Frankenstein’s monster, warning us repeatedly recently, but now are officially primally screaming out for help with a megaphone into the world for all to hear, that we have to at least “pause” this madness.

The new @AnthropicAI post on the approach of recursive self-improvement is very important. It chimes with my argument that some kind of international effort is needed to brake the completely unconstrained AI arms race: https://t.co/kexd0jnra6 pic.twitter.com/JQOrn2Ct3l
— Niall Ferguson (@nfergus) June 5, 2026

Of all people, Mitt Romney totally freaked out about this. In a very non-Mitt way. I don’t blame him.

Our highest and most urgent national priority should be AI safeguards. The risks of AI weapons, pathogens, mass unemployment, surveillance, and even extinction must not continue to be largely ignored. https://t.co/79FzhRMIrq
— Mitt Romney (@MittRomney) June 5, 2026

Magnifica Humanitas/Magnificent Humanity—Pope Leo XIV; The Vatican.

When AI Builds Itself—Marina Favaro and Jack Clark, edited by Santi Ruiz; The Anthropic Institute.

The Real Context News AI Reading List: Anthropic Claude Mythos Glasswing Edition (#7)

Brian E. Frydenborg — Sun, 24 May 2026 16:05:12 +0000

You will be hearing ohhh so much about Anthropic, Claude, Mythos, and Glasswing if you haven’t already. If you don’t know what all these are now, you will need to. Regardless of how well- or poorly-versed you think you are, start here.

(Traduce en español/translate to Spanish)

By Brian E. Frydenborg (Twitter @bfry1981, LinkedIn, Bluesky, Facebook, Substack with exclusive informal content) May 24, 2026; because of YOU, Real Context News surpassed one million content views on January 1, 2023, but I still need your help, please keep sharing my work and consider also donating as I make my overdue comeback! Real Context News produces commissioned content for clients upon request at its discretion. All other editions of the Real Context News AI Reading List can be found here. And be sure to check out my related book review of Nicole Perlroth’s must-read, award-winning book on cyberwarfare as well as my own work on cyberwarfare. And finally, apologies for the delay, so here’s a real meaty edition on a meaty topic!

Technosaurus

Let’s jump into it! Mythos, a particular iteration of the artificial intelligence juggernaut Anthropic’s Claude large language model (LLM) chatbot, is here and doing its thing in the form of Anthropic’s Project Glasswing.

Yes, these tech people are way too in love with their naming schemes, but the bottom line is that Anthropic just announced it has been using its new Mythos version (still in development and often more specifically referred to as Mythos Preview) of the Claude LLM in a massive project with some of the biggest names in tech overall: Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks that has been something of a secret, but is now unveiled as Project Glasswing.

First, let’s start with a few primers on what, exactly, the Mythos (Mythos Preview) edition of Anthropic’s Claude LLM model actually is:

Anthropic’s Mythos Will Force a Cybersecurity Reckoning—Just Not the One You Think—Lily Hay Newman; Wired.
What is Claude Mythos and what risks does it pose?—Liv McMahon Joe Tidy; BBC News.
It’s the End of the Internet as We Know It—Raffi Krikorian (Mozilla CTO); The New York Times.
What is Mythos, Anthropic’s unreleased AI model, and how worried should we be?—Chris Stokel-Walker (edited by Eric Sullivan); Scientific American.
What is Mythos AI and why could it be a threat to global cybersecurity?—Dan Milmo, Kalyeena Makortoff, and Aisha Down; The Guardian.

Anthropic seems to have won a key portion of the AI development race that more and more resembles an arms race, the space race, and the Manhattan Project race (take your dangerous pick of a comparison). It won by developing the Mythos version of its popular Claude LLM model, and Mythos has an advantage over its competition at the time in that Mythos can, and is, finding hacking security vulnerabilities no other model, let alone a human, has been able to yet find. And, according to new information released by Anthropic, it is finding thousands of the dangerous vulnerabilities. In the wrong hands, this can be incredibly destructive, so much so that it is shrouded in secrecy and deemed too dangerous for a general public release by Anthropic. Yes, like Big Tech Santa, Anthropic is keeping its list and constantly automatedly checking it, determining whom it thinks has the ethical and moral substance to be granted access.

Here is the detailed introduction to Glasswing and its first major update on results: Project Glasswing: Securing critical software for the AI era and Project Glasswing: An initial update—Anthropic.
Anthropic sure as hell said a hard no to Secretary of Defense (not legally “War,” sorry fratbros) Pete Hegseth and his Pentagon, which had refused to limit the use of Anthropic’s AI for “mass government surveillance” and “fully-automated weapons,” to quote a detailed explanation from Dario Amodei, Anthropic’s CEO, which I cited in my very first of these AI Reading Lists. That legal fight is still ongoing and is shaping much of the below discussion… Appeals court judges appear to be divided over Pentagon’s legal dispute with AI company Anthropic—Michael Kunzelman, The Associated Press/Federal News Network.
Mozilla’s CTO has a stark warning: It’s the End of the Internet as We Know It—Raffi Krikorian; The New York Times.
BBC isn’t afraid to be afraid with the excellent Scott Galloway: ‘Is Anthropic the new Dr Frankenstein?’—BBC News/AI Decoded.

Tom Friedman sees major warning signs in the Pentagon-Anthropic fight: Anthropic’s Restraint Is a Terrifying Warning Sign—Thomas L. Friedman, The New York Times.
Speaks for itself: The risks of Mythos are no myth, The Editorial Board, Financial Times.
Yes, this is a big deal… Six Reasons Claude Mythos Is an Inflection Point for AI—and Global Security—Gordon M. Goldstein, Council on Foreign Relations.
So much is poised to change as far as cybersecurity: The Detection Era Is Over: How Mythos Changes Everything—Jesse Green; Rubrik.
Umm, don’t forget the workers and watchdogs?? Anthropic Warned Big Companies About Mythos. Workers and Watchdogs Need a Seat at the Table—Amber Scorah, Rebecca Petras; Tech Policy Press.

Not long after telling the Pentagon no, Anthropic also told Xi Jinping and the Chinese Communist Party to take a hike on seeking access to Mythos: China Sought Access to Anthropic’s Newest A.I. The Answer Was No.—Dustin Volz, Julian E. Barnes, Sheera Frenkel, and Tripp Mickle; The New York Times.) Two bad actors, two NOs from Anthropic. Good for Anthropic!
The absurd Pete Hegseth and his flailing Pentagon sure hate Anthropic for clinging to ethics and morals, but still want to use Mythos without Anthropic’s ethics and morals: Pentagon Reportedly Plans to Adopt and Weaponize Latest Cyber-Capable AI Models—Mike Pearl; Gizmodo.
Despite being out with the Pentagon, Anthropic is apparently has been cool with the National Security Agency (NSA) spying agency and will continue to be cool with it: White House Approves $9 Billion for Spy Agencies to Catch Up on A.I.—Dustin Volz and Julian E. Barnes; The New York Times.
Japan’s government and its premier financial institutions are supposedly set to receive access to Mythos, although this is the U.S. government in 2026 so if something completely contradictory happens, don’t be surprised: Japan gov’t to get access to Anthropic’s latest AI: finance minister—Kyodo/Japan Today. This is all in the context of the Japanese government very much taking cybersecurity seriously and proactively in the wake of Mythos’ announcement: Japan scrambles to boost defenses against threats from Mythos—The Asahi Shimbun.
The arrival of mythos has complicated the Trump Administration’s AI “plans” (think more vibes, but dangerous and reckless): How Anthropic’s Mythos Threw the White House AI Strategy Into Chaos—Amrith Ramkumar, Brian Schwartz, and Natalie Andrews; The Wall Street Journal, even as, somehow, the Trump Administration is involved in its rollout and not in a transparent way: White House Opposes Anthropic’s Plan to Expand Access to Mythos Model—Robert McMillan and Amrith Ramkumar; The Wall Street Journal.
This is all in the context of Trump just days ago backing down from an executive order for a major safety review of AI, at the behest of Big Tech: How big tech got its way on Trump’s AI executive order—Nick Robins-Early, The Guardian and Trump calls off AI executive order over concern it could weaken US tech edge—Collin Binkley and Matt O’Brien, The Associated Press.
BONUS: A co-founder of Anthropic, Christopher Olah, is about to launch an encyclical on AI—Magnifica Humanitas—tomorrow with Pope Leo. A primal scream for ethics in an unethical field, if ever there was one… What we know about Pope Leo’s upcoming A.I. encyclical—Colleen Dulle and Gerard O’Connell; America: The Jesuit Review/Inside the Vatican.

Up next? A whole edition of the List on the Elon Musk vs. Sam Altman OpenAI/ChatGPT trial that just concluded. And if you want to spread the word? It’s easy: AI reading list dot com: AIreadinglist.com!