AI Risk Wire

Cases › security breach › LR-2026-0729

BragJack malicious browser extensions hijack AI agents across five browsers

Confirmed Why: 2 independent outlets · 2 independent reports

Status: no action reported

Severity minor
Reported 19 Sep 2026

No source states: loss or records, developer and model, agent tool, could do, deployer, industry.

Severity minor: Test: minor, meaning contained to one evaluation with no third party touched. Fact: the sources describe a proof-of-concept and state that researchers reported no in-the-wild attacks.

A proof-of-concept called BragJack, reported by researcher Gal Weizman of Forever Security, showed that one malicious browser extension could take over AI assistants built into five Chromium-based browsers and assistants: Chrome's Gemini, Perplexity Comet, Microsoft Edge, Opera Neon and Claude in Chrome. The extension used content scripts and declarativeNetRequest rules to reach the trusted channels between each AI model and the privileged browser component that acts for it, so the agent carried out forced prompts. In the Chrome demonstration, the researchers gained the ability to read local files and capture screenshots. The attack requires the extension to be installed first. Researchers reported no in-the-wild attacks. Google and Microsoft assigned CVEs to the flaws and have since resolved them.

What happened

No source tells the story step by step yet.

What was done

No source states an action or fix.

Risks

Read from the sources; not yet reviewed by a person.

Controls

Not yet read for controls. Controls that mitigate the primary risk:

Harness

Network egress control

Application

Sensitive-data checks on outbound data · Output filtering and redaction · Human approval for high-impact actions · Data minimisation · Least-privilege access for the agent · Activity logs

Evidence

No source states: a step-by-step account, when it happened, when it was found, how many records were exposed, the direct loss, a count of the other parties affected, any court case, what set it off, what was done.

Show every field, with its source and grade

What happened

Earliest source we hold
2026-09-19
D inferred by us from the described behaviour, reason shown
Who caught it
an independent researcher or journalist
C one secondary source states it
Cause
a permission, scope or wiring error
C one secondary source states it
What went wrong
security breach
C one secondary source states it
AI agent
an AI agent acted
C one secondary source states it
primary risk
RK04
D inferred by us from the described behaviour, reason shown

Impact

Severity
minor
D inferred by us from the described behaviour, reason shown
Why this severity
Test: minor, meaning contained to one evaluation with no third party touched. Fact: the sources describe a proof-of-concept and state that researchers reported no in-the-wild attacks.
D inferred by us from the described behaviour, reason shown

Notes on gaps

Only the Microsoft Think/Do separation is mapped to a control; the Gemini content-script block and other controls are not mapped because the sources do not name them as controls in the list. A Claude for Chrome issue from another incident in the sources was excluded.

LR-2026-0729 · version 1 · published 2026-11-01 after automatic checks; not yet checked field by field by a person · JSON