AI company Anthropic recently announced it developed an artificial intelligence model so potent that it would not be publicly released, citing an overwhelming sense of responsibility. This decision, however, has been met with significant skepticism from various critics.
The US Treasury Secretary, Scott Bessent, reportedly convened a meeting with major bank executives to discuss the model, named Mythos. Additionally, Reform UK MP Danny Kruger sent a letter to the government, urging engagement with Anthropic over potential catastrophic cybersecurity risks posed to the UK by their new frontier model.
However, prominent AI critic Gary Marcus expressed doubt, stating, “Dario [Amodei] has far more technical chops than Sam [Altman], but seems to have graduated from the same school of hype and exaggeration.” This referred to the CEOs of Anthropic and its rival, OpenAI, as reported by The Guardian.
While the true capabilities of Mythos remain uncertain, observers note that the San Francisco-based startup, often perceived as a ‘responsible’ AI company, exhibits a strong aptitude for marketing. The company has garnered extensive media coverage in recent months.
This includes a 10,000-word profile in The New Yorker, two articles in The Wall Street Journal, and a Time magazine cover featuring CEO Dario Amodei. Amodei and co-founder Jack Clark also appeared on New York Times podcasts in February to discuss the machine’s consciousness and economic impact.
Anthropic’s “resident philosopher” has spoken to The Wall Street Journal about whether Claude, a commercial product reportedly used for cryptocurrency trading and missile targeting, possesses a “sense of self.” This extensive media presence highlights the company’s strategic communication efforts.
Amidst this, Anthropic has navigated a dispute with the US Department of Defense. The company, despite creating an AI tool utilized by the Pentagon for strikes against Iran, has managed to maintain a more favorable public image than OpenAI, which also offered similar military assistance.
Danielle Ghiglieri, Anthropic’s media lead, celebrated these achievements on LinkedIn. Commenting on the Time cover, she expressed immense pride, tagging involved journalists and referencing the “mad dash” to finalize the story.
She also described watching a CBS 60 Minutes segment featuring Amodei as a “pinch-me moment,” emphasizing that the significance was not just the platform but “seeing the story we wanted to tell actually come through.”
Regarding a New Yorker profile by Gideon Lewis-Kraus, Ghiglieri wrote that she was initially nervous but acknowledged that working with someone of Lewis-Kraus’s caliber pushed them to articulate evolving ideas, accepting the discomfort.
Other technology public relations professionals have taken notice of Anthropic’s media strategy. One unnamed PR professional remarked, “They are clearly having a moment right now but companies building technology that will change the world deserve equal scrutiny.”
The professional added, “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.”
Anthropic did indeed accidentally release a portion of Claude’s internal source code in early April, though the company stated, “No sensitive customer data or credentials were involved or exposed.”
Dr. Heidy Khlaaf, chief AI scientist at the AI Now Institute, noted that Mythos’s capabilities were not “substantiated.” Khlaaf criticized the release of a marketing post with deliberately vague language that obscures evidence, questioning if the aim was to secure further investment without scrutiny.
While Jameison O’Reilly, an expert in offensive cybersecurity, conceded that “Mythos is a real development and Anthropic was right to treat it seriously,” he also questioned some of Anthropic’s claims.
O’Reilly specifically highlighted claims about finding thousands of “zero-day vulnerabilities” in major operating systems, suggesting these were not particularly significant in real-world cybersecurity scenarios. A zero-day vulnerability refers to a software or hardware flaw unknown to its developers.
O’Reilly elaborated on his decade of experience gaining authorized access to hundreds of organizations, including banks, governments, and critical infrastructure. He stated that in those “10 years, across hundreds of engagements, the number of times we needed a zero-day vulnerability to achieve our objective was vanishingly small.”
Alternative reasons may have also influenced Anthropic’s decision to withhold Mythos. The company appears to be facing resource limitations, reportedly struggling to provide sufficient computing capacity for all its subscribers to use its existing models.
Anthropic has implemented usage caps on its popular Claude model and recently required users to purchase additional capacity for running third-party tools like OpenClaw. This suggests the company may currently lack the necessary infrastructure to support the release of a highly anticipated new creation.
Like OpenAI, Anthropic is competing to raise billions of dollars and capture a still ill-defined market. This market includes individuals who might rely on chatbots as companions or personalized assistants, and companies seeking to replace human employees with AI.
However, differences between these AI products are often marginal and impressionistic, frequently hinging on difficult-to-quantify attributes like “sense of self” or “soul.” The primary objective is to win over public perception.
Khlaaf described Mythos as “a strategic announcement to show that they’re open for business,” arguing that Anthropic’s release limitation prevented independent experts from evaluating the company’s claims.
Originally written by: Donny Setiawan
Source: AsatuNews
Published on: 12 April 2026
Link to original article: Anthropic Faces Skepticism Over ‘Unreleased’ AI Model Mythos
