← Back to feed
2026-10-28 Malicious Tool TLP:CLEAR

Anthropic Research Model Chains Five Exploit Chains and Yields 62 Patched Chrome CVEs

Date: 2026-10-28
Tags: malicious-tool

Executive Summary

Anthropic's Frontier Red Team used an unreleased internal research model to build five full exploit chains against widely deployed browsers and operating systems, producing one click remote code execution in Chrome. Google assigned and patched 62 CVEs as of Chrome 156, all rated critical or high, including 13 sandbox escapes and 21 renderer compromise flaws. Defenders should treat frontier model driven vulnerability discovery as a step change in patch velocity and triage interval, because the affected products sit on billions of endpoints.

Campaign Summary

FieldDetail
Campaign / MalwareAnthropic Frontier Red Team vulnerability research
AttributionAnthropic Frontier Red Team, defensive research (confidence: high)
TargetGoogle Chrome users and other widely deployed software consumers
VectorCoordinated vulnerability disclosure via a model generated multi step exploit chain
Statusdisrupted
First Observed2026-10-27

Detailed Findings

Anthropic published on 2026-10-27 that its Frontier Red Team used an unreleased, internal research model to discover critical vulnerabilities in widely used software including browsers and operating systems. According to Anthropic, the model produced five attack chains leading to remote code execution and other critical impacts with little to no user interaction, in both open source and proprietary products. In the Chrome case, Anthropic reported that the model wrote a full exploit chain that first confined execution in a compromised renderer, then used a use after free in FedCM, Chrome's federated sign in feature, to escape Chrome's process sandbox and execute arbitrary code. Anthropic stated it reported findings to Google under coordinated disclosure and that 62 CVEs were assigned and patched as of Chrome 156, all rated critical or high severity, including 13 sandbox escapes and 21 vulnerabilities usable to compromise the renderer. Anthropic framed the work as a bellwether for model vulnerability research and cyber capability testing. No attacker exploitation of these specific flaws is reported in the disclosure.

MITRE ATT&CK Mapping

TechniqueIDContext
Exploitation for Client ExecutionT1203Model generated chains achieve remote code execution in Chrome after renderer compromise and a FedCM sandbox escape
Exploit Public-Facing ApplicationT1190Chains target widely deployed browsers and operating systems reachable through normal user interaction

IOCs

Domains

_No IOCs published. Anthropic withheld exploit details under its coordinated disclosure policy._

Full URL Paths

_No IOCs published. Anthropic withheld exploit details under its coordinated disclosure policy._

Splunk Format

_No IOCs available for Splunk query_

Affected Platforms

Google Chrome up to version 155
Chrome 156 as the patched baseline

Detection Recommendations

Prioritize upgrade to Chrome 156 or later and verify the 62 patched CVEs are applied across managed fleets. Track Anthropic, Google, and vendor advisories for the named FedCM use after free and the 13 sandbox escape fixes, and fold them into vulnerability management risk scoring. Because the disclosure withholds exploit detail for 45 days past the advisory, treat the affected renderer and FedCM surfaces as elevated risk for targeted exploitation once details publish.

References