The Guardian just published the single most devastating analysis of Anthropic marketing strategy to date. And it did not come from some anonymous blogger or a jealous competitor. It came from cybersecurity researchers, AI policy experts, and PR professionals who all arrived at the same conclusion: Anthropic is running the same playbook OpenAI used, and it is doing it better.

The piece, titled "Too powerful for the public," methodically lays out how Anthropic has turned Mythos into the most effective product launch in AI history. Not by releasing the model. By refusing to release it.

The receipts

In the past few months, Anthropic has secured a 10,000-word New Yorker profile, two Wall Street Journal features, the cover of Time magazine (Dario Amodei towering above the Pentagon), two separate New York Times podcast appearances, and a CBS 60 Minutes segment. The company media lead, Danielle Ghiglieri, has been openly celebrating the wins on LinkedIn, calling them "pinch-me moments" and praising the "story we wanted to tell."

One tech PR professional told The Guardian: "They accidentally leaked their own source code last week, then this week they claim stewardship over cyber threats with a new powerful model that only they control. Any other big tech firm would be ridiculed."

That is the paradox nobody in the mainstream coverage has confronted until now. Anthropic bungled its own code security, then told the world it had built something too powerful to be trusted in anyone else hands. And the media ate it up.

The zero-day claims do not hold up

The most explosive claim from Anthropic was that Mythos discovered "thousands of zero-day vulnerabilities" in major operating systems, browsers, and widely used software. That sounds terrifying. The problem: offensive cybersecurity experts say it is not that significant.

Jameison O Reilly, who has spent over a decade gaining authorized access to hundreds of organizations including banks, governments, and critical infrastructure, told The Guardian that in all those engagements, "the number of times we needed a zero-day vulnerability to achieve our objective was vanishingly small." Most real-world attacks do not require zero-days. They exploit misconfigurations, weak credentials, and human error.

Dr. Heidy Khlaaf, chief AI scientist at the AI Now Institute, was more direct: Anthropic claims were not "substantiated." She called the Mythos announcement "a marketing post with purposely vague language that obscures evidence" and questioned whether the goal was to "garner further investment without scrutiny."

The bait and switch

Here is the quote that should make every Anthropic supporter uncomfortable. Khlaaf suggested we may be "seeing the very same bait and switch playbook that was used by OpenAI, where safety is a PR tool to gain public trust before profits are prioritised." And then the kicker: "Anthropic publicity has managed to better obscure this switch than its rivals."

The Guardian also raised a practical explanation for why Anthropic is not releasing Mythos: it might simply not have the infrastructure. The company has been struggling to provide enough compute for existing Claude users, introducing usage caps and forcing subscribers to pay extra for third-party tool access. Launching a new model it has hyped as the most dangerous AI ever built might be something Anthropic literally cannot afford to do.

My take

I have been covering the Mythos story since Anthropic first teased it. I wrote about the Treasury Secretary emergency meeting. I wrote about the Bank of England summoning British banks. I reported on multiple governments scrambling to assess the risk. And I would have continued treating this as a pure safety story if not for The Guardian doing what journalists are supposed to do: asking the uncomfortable questions.

Mythos may genuinely be a powerful model. Anthropic may genuinely believe it poses cybersecurity risks. But the media strategy around it has been flawless in a way that should make everyone suspicious. When a company that leaked its own source code two weeks ago tells the world it has built something too dangerous to release, and then rides the resulting panic to a Time magazine cover, a CBS segment, and a $380 billion IPO valuation, you have to ask: is this safety, or is this salesmanship?

The answer, most likely, is both. And that is what makes it so effective.