Anthropic (AI company)

Anthropic

Anthropic

new

xīn

新

model

móxíng

模型

Mythos (AI model name)

Mythos

Mythos

to discover / to uncover

fāxiàn

发现

tens of thousands

shù wàn

数万

measure word

gè

个

network / cyber

wǎngluò

网络

security / safety

ānquán

安全

vulnerability / security flaw

lòudòng

漏洞

。

New Words

News in English

Anthropic Mythos AI Uncovers Tens of Thousands of Cybersecurity Vulnerabilities

Anthropic AI model Mythos has discovered tens of thousands of security vulnerabilities across major operating systems and browsers, raising alarm in the cybersecurity industry.

Anthropic's new generation AI model Mythos recently identified tens of thousands of zero-day vulnerabilities across mainstream operating systems and browsers. In the Firefox browser alone, the previous Anthropic model found approximately 20 vulnerabilities, while Mythos discovered nearly 300. This model is equally efficient in patching vulnerabilities, but this also means it can equally exploit vulnerabilities to launch attacks. Anthropic has restricted Mythos access permissions to a small number of United States enterprises, including Apple, Amazon, and JPMorgan Chase. Additionally, Anthropic commits to provide up to 100 million dollars in usage quota, supporting open-source security communities to address related threats. Industry experts warn that currently corporate vulnerability patching still requires several days to several weeks, while AI vulnerability discovery speed is far faster than this, making security risks intensifying. The United States government has begun to re-examine regulatory policies for frontier AI models.

More in Science & Technology