Anthropic’s cybersecurity platform Mythos is beginning to change how governments and organizations think about digital defense—and now the artificial intelligence firm is loosening its grip on the information generated by the system.
Anthropic has kept Mythos under tight wraps due to the potential for abuse—the model’s capability to detect vulnerabilities could be devasting in the hands of cybercriminals. Until now, Mythos has only been deployed to a handful of tech firms and the U.S. government, which have been exploring the system’s capabilities.
However, despite Mythos’ unprecedented cybersecurity capabilities, it would not be fully effective if deployed solely in siloes. Too often, modern fraud incidents are not isolated events. Instead, they are frequently part of organized campaigns that also leverages AI, without any guardrails.
This threat has heightened the need for shared intelligence: industry data that organizations can use to properly categorize, identify, and report threats. As a result, Anthropic has told its partners they can share findings, best practices, and tools built with Mythos. And not just with one another, but also with other companies, agencies, media outlets, and even the public.
The Looming Surge of Fraud Attacks
Mythos is already prompting companies and even governments to rethink their approach to cybersecurity. For example, Japan called for a consortium to examine the global consequences of AI systems that could potentially exploit zero-day vulnerabilities—software flaws unknown to developers—faster than organizations can patch them.
More recently, Palo Alto Networks warned that a significant surge in AI-powered fraud attacks is looming and that organizations may have only months to strengthen their cybersecurity defenses.
Losing Faith in Defenses
This anticipated spike in AI-driven fraud is why the International Monetary Fund recently urged financial institutions not only to enhance their infrastructure, but also to share data—something highly regulated banks have long been reluctant to do. However, the IMF warned that the fraud landscape has become too sophisticated and fast-moving for financial institutions to continue operating in isolation.
This creates financial and infrastructure risks for banks and other institutions, but it is not the only consequence of widespread AI-enabled fraud. Even in the current environment, consumers are already beginning to lose confidence in institutions’ ability to protect them.
