Researchers report code-execution and file-theft flaws in the browsers built into five AI coding assistants

Cover illustration for Forever Security's EMBRACE research on embedded browsers in AI assistants. — Forever Security

Forever Security calls the shared root cause EMBRACE, an embedded browser trusted like a real one across Cursor, GitHub Copilot, Codex, VS Code and Claude Desktop; it says the Codex and VS Code issues were unpatched at publication.

Read at Forever Security

+ Help Net Security

More from Snape Report

Tech stocks lose more than $1 trillion as Intel falls 29%, its worst single-day drop ever
Mistral releases open weights for Mistral Large 4, its 1-trillion-parameter model
Amazon discloses a secured loan facility of up to about $80 billion to buy AI infrastructure hardware
Japan asks Anthropic and OpenAI for wider access to advanced cybersecurity models, Nikkei reports
Google's Suncatcher satellite returns the first AI response from a data-centre TPU in orbit
OpenAI reorganizes its Model Spec around five core values
Tesla is expected to open Optimus robot reservations in early December, the San Francisco Standard reports
Google, Microsoft, OpenAI and Anthropic back a shared robots.txt signal for opting out of AI training

Front page →