"AI models have reached a level of coding capability where they can surpass all but the most skilled humans at finding and exploiting software vulnerabilities."
"Engineers at Anthropic with no formal security training have asked Mythos Preview to find remote code execution vulnerabilities overnight, and woken up the following morning to a complete, working exploit."
"Mythos Preview generated a working exploit 72.4 percent of the time, marking a significant advancement in AI's ability to identify and exploit software vulnerabilities."
"Anthropic chose to provide a preview version to a set of industry partners so they can use it to find flaws in their systems before adversaries do."
Anthropic developed an AI model named Mythos that can generate zero-day vulnerabilities with a success rate of 72.4%. Unlike its predecessor, Claude Opus 4.6, which had a negligible success rate, Mythos can produce working exploits quickly. Anthropic decided against releasing Mythos publicly, fearing it would cause significant harm. Instead, they initiated Project Glasswing, providing a preview version to select industry partners to identify vulnerabilities in their systems, supported by substantial funding for open-source security initiatives.
Read at Theregister
Unable to calculate read time
Collection
[
|
...
]