Anthropic's Mythos AI Uncovers Over 2,000 Software Vulnerabilities, Raising Cybersecurity Concerns
Anthropic has developed a new AI model named Mythos, designed for defensive cybersecurity research. During a seven-week testing period, Mythos identified over 2,000 previously unknown software vulnerabilities, a significant portion of the annual global output of zero-day vulnerabilities. This discovery highlights the model's capability to expose hidden weaknesses in software that have been overlooked by human security researchers for decades. Due to the potential risks associated with its capabilities, Anthropic has decided to restrict access to Mythos, allowing only a select group of trusted partners, such as Microsoft and Google, to experiment with it under controlled conditions. This decision underscores the model's power and the need for careful consideration of its implications. 
