It may be apt that Anthropic’s headline-grabbing new AI model is called Claude Mythos. After all, the name comes from the Greek word for word or story, and the AI giant sure is telling a tale about Mythos’ capabilities. You could call it good marketing as Anthropic aims to turn from AI vendor into a household big tech name like Google. Funny thing is, though, Google is also spinning a yarn or two, and in the same security wheelhouse as the incumbent. The AI giants have turned apocalyptic, if you’ll excuse my own hyperbole.

First, a recap: as you’ve probably heard, Anthropic’s Claude Mythos Preview is reportedly its most powerful large language model (LLM) to date. Although a general-purpose model, Mythos' launch fanfare centered around its particular knack for computer security tasks like vulnerability discovery. This led to Anthropic forming Project Glasswing, a cybersecurity venture with partners like Nvidia, Palo Alto Networks, Broadcom, Cato Networks, Cisco, and, yes, Google, to test Claude agents powered by the model in finding vulnerabilities on their systems – some of which lay uncovered by humans for three decades until AI came along.

That news eclipsed an announcement from the week before, when Google set 2029 as the deadline for post-quantum cryptography (PQC) migration. According to the hyperscaler, the world is on the cusp of a quantum computer emerging and thus breaking today's encryption with potentially apocalyptic results.

Both announcements certainly got people talking. The Google news moved the quantum apocalypse needle from somewhere around 2035 to this decade, with Cloudflare and others fast-tracking their own quantum roadmaps to the same deadline.

Anthropic’s claims, meanwhile, led to portents of extinction for today’s security vendors. After all, if Mythos is as good as Anthropic alleges, then who needs a pricey software subscription when you can just use the latest iteration of Claude to find bugs in your network?

Some tech titles even looked at the bigger picture, wondering just how powerful Mythos is in general. If it’s more than a cybersec aid, what else is it good at, and is it good enough to pose an extinction-level event (ELE) for any other professions? Or even worse, will it help attackers find countless new flaws to exploit, thus bringing the world to a standstill?

To use another word Greek word, Anthropic’s release is more "peitho" than mythos, meaning persuasion in shaping opinion and policy. The masterstroke was Anthropic claiming it won’t make Mythos publicly available for fears it may fall into the wrong hands. Hence it being limited to its Glasswing partners, with detriment to any cybersecurity players not on that exclusive list. That only made the ELE paranoia worse for vendors who didn't get the VIP invite.

The two gatekeepers of the apocalypse

Google is also guilty of gatekeeping, as much as it is of more good marketing. Seeing the reactions of the quantum computing community, the general consensus was that Google buried the quantum lede somewhat by tying its Q-Day mast to the cryptocurrency flag, with a blog on safeguarding cryptocurrency by disclosing quantum-related vulnerabilities “responsibly.”

It’s as if Google thought the tech bro audience is more interested in their crypto stock than when the first quantum computer comes online. While some deemed the publicity move as cringey, to be honest, who could blame any mainstream tech enthusiasts after years of hearing Q-Day is only “five years away” for, what, almost two decades now?

What pushed Google to this point was its own research, with claims it’s decreased the number of qubits needed to break 2,048-bit RSA encryption, dropping from 20 million to one million, as well as its Willow chip reportedly able to use error correction in a way that improves reliability as it scales up, instead of the opposite.

It seems little to none of this has been independently verified, and according to insiders, that’s how it’s going to be going forward. Where the quantum community was once open in its efforts to make quantum computing a reality, now that we’re in the final stages of actualizing cryptographically-relevant quantum computers, the big names are closing ranks and keeping its tech proprietary. Expect a lot of what happens next to happen behind closed doors, much like the work on Project Glasswing (of which Google is also a member, remember).

And expect a lot of hype and "holier than thou" vibes, even if whatever happens behind the scenes is far from the truth. As David Girvin, AI security researcher at Sumo Logic told this title regarding Project Glasswing: “Of course it’s clever marketing. ‘We built something incredible but you can’t touch it yet’ is one of Silicon Valley’s oldest confidence plays. It gets headlines, drives fear of missing out (FOMO), and positions the company as the adult in the room that’s thinking deeply about safety while everyone else is pushing half-baked AI agents into production.”

All we can do is enjoy the ride. After all, how's that song go? "Open the door, get on the floor – everybody walk the dinosaur!"