A newly revealed AI tool, dubbed Claude Mythos Preview, has been developed by Anthropic and is reported to possess capabilities that could potentially compromise a vast number of computer servers globally. This model is said to have identified thousands of cybersecurity vulnerabilities across major operating systems and browsers, a level of access typically reserved for state-sponsored hacking entities from nations like China, Russia, and the United States.
On Tuesday, Anthropic disclosed the existence of this powerful bot, which will initially be accessible only to a consortium of leading technology firms, including Apple, Microsoft, Google, and Nvidia. These companies plan to utilize Mythos Preview to enhance the security of their software by identifying and addressing bugs and exploits. Anthropic has opted against a public release for now, citing concerns over safety without adequate safeguards.
Cybersecurity professionals have long cautioned about the potential threats posed by advanced hacking bots. The emergence of AI models capable of coding has heightened the risks, as these systems can efficiently uncover software vulnerabilities. The introduction of Mythos Preview could signify a significant shift in cyberattack capabilities, allowing unprecedented scale and speed in exploiting weaknesses in digital infrastructure.