From BigGo Finance · BigGo Finance · · 1m 30s
Anthropic Unreleased AI Model Creates Exploits
Anthropic's internal AI model chained zero-days into working exploits across Chrome, Safari, Android, Windows and Linux.
In brief
Anthropic's internal AI model chained zero-days into working exploits across Chrome, Safari, Android, Windows and Linux. Anthropic's unreleased AI model identified and chained vulnerabilities in major platforms, creating working exploits that have been patched by vendors. Originally reported by BigGo Finance.
Anthropic Discloses AI Model Exploits
Anthropic said an unreleased internal AI model identified previously unknown vulnerabilities in Chrome, Safari, Android, Windows and Linux and chained them into working exploits in laboratory tests.
Chrome Exploit Chain Details
The Chrome chain combined four flaws to escape the browser sandbox, while the Safari attack disabled the same-origin policy.
Android Exploit Requires Physical Access
An Android exploit required physical USB access to extract photos and data from a locked phone.
Windows Exploit Causes Server Crashes
A Windows chain caused repeated server crashes through HTTP/3 request handling without executing code.
Linux Exploit Grants Root Access
A Linux chain achieved root access over the network under specific conditions without user interaction.
Vendors Patched Flaws
Anthropic reported all flaws to vendors, which have issued patches; Apple and Google credit the company and its Claude model in release notes.
Public Records Don't Fully Match
For Windows and Linux chains, public vulnerability records do not fully align with Anthropic's descriptions.
Anthropic's Defense Strategy
Anthropic framed the research as a defensive contribution with a coordinated disclosure policy giving developers 90 days to patch.
At a glance
Anthropic's unreleased AI model identified and chained vulnerabilities in major platforms, creating working exploits that have been patched by vendors.
