"Anthropic's new AI model, Mythos, is so skilled at finding and exploiting software vulnerabilities that it's too dangerous to release. The company is limiting access to a small group of major technology companies."
"Researchers are concerned about both the scale and accessibility of AI‑enabled attacks. Tasks that once required advanced expertise are increasingly being automated or semi‑automated by AI systems."
"Even without the powerful new model, existing, publicly available AI models can already carry out sophisticated cyberattacks-sometimes in minutes, raising questions for enterprises and policymakers."
Anthropic's new AI model, Mythos, is designed to find and exploit software vulnerabilities, prompting concerns among cybersecurity experts. The company is limiting access to major tech firms to allow time for system strengthening. Other AI labs, like OpenAI, are developing similar models with phased rollouts to partners. Analysts debate the motivations behind these limited releases, but there is consensus that AI-driven cyber capabilities have reached a critical point, enabling even less skilled attackers to execute sophisticated cyberattacks rapidly.
Read at Fortune
Unable to calculate read time
Collection
[
|
...
]