Anthropic’s ‘Claude Mythos’ model sparks fears of – Business News
Anthropic has triggered alarm bells by touting the terrifying capabilities of “Claude Mythos” – with executives warning that the new AI model is so harmful it could trigger a wave of catastrophic hacks and terror assaults if launched to the broader public.
In a nightmarish evaluation, Anthropic itself revealed that Mythos – if it fell into the mistaken fingers – may simply exploit essential infrastructure like electric grids, energy plants and hospitals. The model has already “found thousands of high-severity vulnerabilities, including some in every major operating system and web browser,” in keeping with the AI company.
Critics have accused Anthropic CEO Dario Amodei of utilizing security fears to advertise the company’s merchandise. Bloomberg by way of Getty Images
Rather than a huge release, Anthropic, led by CEO Dario Amodei, has unveiled “Project Glasswing,” a plan to supply the model to a handpicked group of about 40 corporations, together with Amazon, Google, Apple, Nvidia, CrowdStrike, and JPMorgan Chase, which can obtain early entry to Mythos to allow them to use it to search out and repair security flaws.
The corporate-only rollout is probably going Anthropic’s best doable approach to “give it to the guys to patch the holes, but not to the hackers that are going to find more holes,” Roman Yampolskiy, an AI security researcher on the University of Louisville, advised The Post.
“Most likely, of course, there’s going to be a leakage of some kind,” he stated. “Any degree of restriction is most popular over full open entry. Ideally, I’d like to see this not developed within the first place. And it’s not like they’re going to stop.
“That’s exactly what we expect from those models – they’re going to become better at developing hacking tools, biological weapons, chemical weapons, novel weapons we can’t even envision,” Yampolskiy added.
In one occasion detailed in Anthropic’s testing, Mythos broke out of a secure “sandbox” meant to limit web entry – with a researcher solely discovering out “by receiving an unexpected email from the model while eating a sandwich in a park.” In one other case, Mythos discovered a flaw within the OpenBSD working system that had been hidden in plain sight for 27 years.
Despite the dangers, Anthropic argues Project Glasswing will help the US’ defensive capabilities as adversaries in Iran, China and Russia grow to be ever more aggressive about concentrating on essential infrastructure.
An Anthropic official stated the company “centered on organizations whose software program represents the biggest share of the world’s shared cyberattack floor.
“These are the companies that build and maintain the operating systems, browsers, cloud platforms, and financial infrastructure that billions of people rely on every day,” the official stated. “When you find a vulnerability in one of their systems and it gets patched, that patch protects everyone who uses that software — in many cases, hundreds of millions of people.”
About 40 corporations are collaborating in Project Glasswing. NurPhoto by way of Getty Images
Anthropic stated it’s in lively discussions with US authorities officers about how Mythos can assist the nation’s cyber capabilities — each offensive and defensive.
“Claude Mythos Preview demonstrates what is now possible for defenders at scale, and adversaries will inevitably look to exploit the same capabilities,” stated Elia Zaitsev, chief technology officer at CrowdStrike.
While Mythos seems to be a main leap ahead technologically, critics are unsure about whether or not Anthropic’s actions – together with the splashy public announcement – match its rhetoric concerning the dangers.
CEO of Anthropic Dario Amodei, addresses the gathering on the AI Impact Summit, in New Delhi, India, February 19, 2026. REUTERS
Perry Metzger, chairman of Alliance for the Future, a Washington, DC-based AI coverage group, famous that the hype about Mythos as a product has “spread like wildfire” as a consequence of the company’s warning.
“You’d better carefully pay for access to Glasswing or get in on it, because only they are responsible enough to decide who should and shouldn’t have access. They’re the experts, after all,” Metzger sarcastically stated. “I find the whole thing maddening.”
As The Post has reported, Anthropic’s critics, together with President Trump’s AI adviser David Sacks and others within the White House, have claimed the company’s security warnings are literally an elaborate attempt at “regulatory capture” – Silicon Valley lingo for crafting the foundations in such a method that they benefit and their rivals wrestle.
“At every point in the conversation around the emergence of AI, Dario Amodei believes he, and he alone is qualified to determine what this technology can do, who can access it,” stated Nathan Leamer, govt director of the advocacy group Build American AI. “He alone is the modern Solomon who will decide what entity will regulate it. Who needs the public square?”
Claude Mythos is Anthropic’s newest AI model. Christopher Sadowski for NY Post
The Anthropic official pushed back on claims that Project Glasswing was an attempt at regulatory seize, noting that individuals embody “several of the largest companies in the world, including those that build their own AI models.”
“We made our most capable model available to AWS, Apple, Google, Microsoft, and others to find and fix vulnerabilities in their own systems, and prioritized open-source maintainers, donating $4 million to organizations like the Linux Foundation and Apache Software Foundation,” the official stated.
Another tech industry insider, who spoke to The Post on situation of anonymity, identified that OpenAI equally warned in 2019 that GPT-2, an early model of its chatbot model, was too harmful for release.
Amodei and Anthropic’s high coverage govt Jack Clark have been each working at OpenAI on the time.
The insider speculated that the restricted release may have more to do with Anthropic’s wrestle to keep tempo with compute demand than security fears.
“They are trying to deflect from the fact that they can’t serve the model because they have no compute,” the individual stated.
