Mythos, Anthropic’s most recent AI technology, is so effective at identifying software flaws that the business is hesitant to make the model public for fear that it would end up in the wrong hands.
In a post on its website this week, the company that created the Claude AI chatbot stated that the new tool has already found thousands of vulnerabilities in “every major operating system and web browser.”
While such feature would be helpful for safeguarding vital systems, there are worries that hackers might use Mythos to target the IT infrastructure of banks, hospitals, government agencies, and numerous other institutions.
Getting ready for the “storm”
Instead of making Mythos available to the general public, Anthropic is sharing the technology with a few key businesses, such as Amazon, Apple, Cisco, JPMorgan Chase, and Nvidia, so they can test the model and fortify their own systems against cyberattacks. According to Anthropic, the initiative, known as Project Glasswing, aims to assist important businesses in strengthening their defenses before hackers gain access to Mythos or comparable AI models.
However, security experts noted that the worries around Mythos demonstrate the risks of using AI as a weapon.The CEO of cybersecurity AI firm Assail, Alissa Valentina Knight, told, “We need to take this as a wake-up call to say, the storm isn’t coming—the storm is here.” “We must get ready because when humans hacked into our networks, we were unable to keep up with the adversaries. If they’re utilizing AI, we can’t stay up because it’s so much faster and more powerful.
Federal officials are likewise worried about Mythos’ capabilities. Top bank CEOs met behind closed doors on Tuesday with Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell to discuss Mythos and other new cybersecurity threats arising from artificial intelligence. According to the report, Anthropic also briefed important industry stakeholders and senior U.S. government officials about Mythos’s potential.
In a separate interview that will run on “Face the Nation with Margaret Brennan” on Sunday, IMF Managing Director Kristalina Georgieva stated that the world is unable “to protect the international monetary system against massive cyber risks.”According to Georgieva, “the risks have been growing exponentially.” “Yes, we are worried. We are eager to see greater focus on the safeguards required to maintain financial stability in the AI industry.
A request for comment was not answered by Anthropic. However, the business emphasized the dangers of abusing technologies like Mythos in their statement. “The fallout — for economies, public safety, and national security — could be severe,” the business added.
The weakest part
Hackers already have access to sophisticated AI models and are exploiting them for a variety of nefarious objectives, such as creating autonomous “agents” that can carry out attacks without human intervention. This alarming fact is concealed by such stern warnings.
According to cybersecurity experts, these attacks include anything from creating deepfake videos and initiating ransomware attacks to distributing malware and carrying out identity theft frauds.
In response to a new analysis from PwC, advanced adversaries are utilizing AI to improve precision, scale automation, and shorten attack timelines, while even low-skilled threat actors can now carry out high-speed, high-volume operations due to AI-enabled tooling.
The management consulting firm stated, “In 2025, the time between an AI company’s public release of a new capability and its weaponization by threat actors shrank dramatically, a trend we assess will likely accelerate in 2026.”
Other AI technologies are already increasing the hazards to consumers, companies, and governments, even though they are not yet as successful as Mythos in revealing the soft underbelly of software. For example, according to Zach Lewis, chief information officer at the University of Health Sciences and Pharmacy in St. Louis, hackers are using AI to hone so-called phishing assaults meant to extract private information.
According to him, it has been utilized to truly script those chats, dialogues, and phishing emails for particular individuals. They are then customized to make them much harder to see and determine whether or not they are fraudulent.
Lewis stated, “We’re going to see a lot more vulnerabilities and probably a lot more attacks once [Mythos] drops.” Cyberattacks will undoubtedly rise unless we are able to patch all of those vulnerabilities practically instantly.
According to Knight, AI is more efficient than humans at identifying software vulnerabilities because it can swiftly analyze thousands of lines of code and identify issues, something that people aren’t always skilled at.The weakest link in security is humans, according to Knight. “When developing code, humans are capable of making mistakes. Vulnerabilities in source code might exist that humans have never discovered.
On brand for Anthropic?
Some security experts questioned Anthropic’s stepwise approach of releasing Mythos, claiming that the limited release was intended to spark the interest of additional potential customers.
Meanwhile, the Wall Street Journal reports that both Anthropic and rival OpenAI are slated to pursue initial public offerings before the end of the year, which could be an incentive to generate headlines, according to Peter Garraghan, founder and Chief Science Officer of Mindgard, an AI security platform.
“I suspect Anthropic may be using this as a marketing ploy, perhaps towards IPO,” he stated.
By openly stressing AI safety and showcasing its safeguards for keeping the technology in check, Anthropic has attempted to set itself apart from OpenAI and other competitors. According to Malek Ben Sliman, a marketing lecturer at Columbia Business School, Anthropic’s choice to postpone the release of Mythos and initiate Project Glasswing is consistent with that image.Anthropic has truly remained loyal to its principles in the face of difficult choices,” he stated. Curating the unveiling of Mythos “does allow them to look to be the protectors of this responsible AI, but it also is a great marketing and advertising tool.”
