"Anthropic's Claude Mythos has been reported to outperform humans in various hacking and cybersecurity tasks, raising alarms among regulators and financial institutions about its potential risks."
"The AI model, part of the Claude system, has shown striking capabilities in identifying dormant bugs in decades-old code, which could pose significant security threats."
"Access to Mythos has been restricted to 12 tech companies through Project Glasswing, aimed at enhancing the security of critical software and mitigating risks associated with the AI."
"Researchers involved in testing AI models noted that Mythos could easily exploit vulnerabilities, emphasizing the need for careful consideration of its deployment in cybersecurity."
Anthropic's new AI model, Claude Mythos, reportedly outperforms humans in certain hacking and cybersecurity tasks, prompting concern among regulators and financial institutions. The model is part of a broader AI system called Claude, which competes with OpenAI's ChatGPT and Google's Gemini. To mitigate potential risks, Anthropic has limited access to Mythos through Project Glasswing, involving 12 tech companies, including Amazon, Apple, Microsoft, and Nvidia. Researchers noted Mythos's ability to identify and exploit vulnerabilities in legacy code, highlighting its capabilities in computer security.
Read at www.bbc.com
Unable to calculate read time
Collection
[
|
...
]