"Claude Mythos Preview, an AI model from Anthropic, is considered too dangerous to release publicly due to its exceptional cybersecurity and cyber-attacking capabilities, which could enable widespread disruption."
"Mythos has discovered vulnerabilities in every major browser and operating system, including a 27-year-old bug in critical security infrastructure, raising alarms about potential catastrophic implications."
"Cyber-attacks are evolving from digital threats to real-world consequences, affecting essential services like airports and hospitals, with the new AI model potentially empowering amateurs to execute sophisticated attacks."
"Experts warn that the capabilities of AI have crossed a threshold, fundamentally changing the urgency required to protect critical infrastructure from emerging cyber threats."
In June 2024, a cyber-attack on a pathology services company disrupted London's hospitals, leading to over 10,000 cancelled appointments and a patient's death. Anthropic's AI model, Claude Mythos Preview, is deemed too dangerous for public release due to its advanced cyber-attack capabilities. It has identified vulnerabilities in major browsers and operating systems, including a 27-year-old bug in critical security infrastructure. If widely available, this technology could enable amateurs to execute large-scale cyber-attacks, threatening essential services and systems worldwide, prompting urgent calls for enhanced cybersecurity measures.
Read at www.theguardian.com
Unable to calculate read time
Collection
[
|
...
]