Dopious
Senior Member
Founding Member
Sapphire Member
Patron
Hot Rod
Earlier in April, Anthropic also made headlines in newspapers with Claude Mythos – a model that is a better hacker than almost all humans. Mythos can find vulnerabilities in programs and operating systems so they can be fixed. Anthropic warned at the same time that similar models could instead be used for attacks.
According to Mozilla, an early version of Claude Mythos managed to find 271 vulnerabilities in the Firefox browser, which have been patched in Firefox 150. The developer mentions a sense of vertigo at the realization that Mythos found so many security issues. How will they have time to fix everything?
Mozilla writes that thanks to Mythos, “defenders” finally have a chance to win a decisive victory against attackers. According to Mozilla, the industry previously aimed for a draw. It was unrealistic to release a program with zero exploits – instead, the goal was to make attacks so difficult and expensive that only actors with unlimited budgets could afford to exploit the gaps.
Software security has been dominated by attackers, according to Mozilla, because attackers have an advantage: they only need to find a single chink in the armor or a single rusty link in the long chain. Thanks to Claude Mythos, Mozilla believes the playing field has now changed. The post points out that none of the vulnerabilities Mythos discovered could not have been found by an elite human researcher.
Source: https://blog.mozilla.org/en/firefox/ai-security-zero-day-vulnerabilities/
According to Mozilla, an early version of Claude Mythos managed to find 271 vulnerabilities in the Firefox browser, which have been patched in Firefox 150. The developer mentions a sense of vertigo at the realization that Mythos found so many security issues. How will they have time to fix everything?
Mozilla writes that thanks to Mythos, “defenders” finally have a chance to win a decisive victory against attackers. According to Mozilla, the industry previously aimed for a draw. It was unrealistic to release a program with zero exploits – instead, the goal was to make attacks so difficult and expensive that only actors with unlimited budgets could afford to exploit the gaps.
Software security has been dominated by attackers, according to Mozilla, because attackers have an advantage: they only need to find a single chink in the armor or a single rusty link in the long chain. Thanks to Claude Mythos, Mozilla believes the playing field has now changed. The post points out that none of the vulnerabilities Mythos discovered could not have been found by an elite human researcher.
Source: https://blog.mozilla.org/en/firefox/ai-security-zero-day-vulnerabilities/