"Mozilla's collaboration with Anthropic began with Claude Opus 4.6, which scanned nearly 6,000 C++ files and produced 112 reports, confirming 22 security-sensitive bugs in Firefox 148."
"Bobby Holley, Firefox's CTO, remarked that the experience of using Mythos to identify vulnerabilities was so overwhelming that it gave the team 'vertigo.'"
"The UK AI Security Institute confirmed that Mythos can execute autonomous multi-stage network attacks, highlighting the dual-use tension as a central policy question."
"Mozilla stated that the bugs identified by Mythos were not exotic, emphasizing that no human team could have found 271 vulnerabilities as quickly as the AI model did."
Mozilla released Firefox 150, addressing 271 security vulnerabilities identified by Anthropic's AI model, Mythos. This collaboration began with Claude Opus 4.6, which found 22 bugs in Firefox 148. Mythos significantly outperformed its predecessor, identifying over twelve times as many vulnerabilities. Firefox CTO Bobby Holley noted that while the defects are finite, the speed of detection by Mythos surpasses human capabilities. The UK AI Security Institute confirmed Mythos's ability to execute complex network attacks, raising important policy questions regarding dual-use technology in AI.
Read at TNW | Anthropic
Unable to calculate read time
Collection
[
|
...
]