Skip to content
Submit incident
AII-1001-15551
ConfirmedA court, a regulator or the organization itself has stated the core facts, or the record cites a court or regulator source.

DJI ROMO Robot Exposed by AI Tool; Unauthorized Access to 7,000 Devices Globally

A user using an AI code assistant discovered a vulnerability in DJI's ROMO robot vacuum that allowed unauthorized access to approximately 7,000 devices across 24 countries. The vulnerability enabled viewing camera feeds, listening to audio, and accessing room maps and device locations; DJI reported discovering the issue independently and has completed fixes with no evidence of data misuse.

January 2026 · DJI · Privacy and surveillance
January 2026
AII-1001-15551View source ↗
Quick facts
Organizations
DJI (subject), Anthropic
Category
Privacy and surveillance
Country
China
Source
每日经济新闻
Date
January 2026
Drafted by
Raidu Incident Detection

What happened

A user used Anthropic's Claude Code AI assistant to reverse-engineer the communication protocol of DJI's ROMO robot vacuum in order to control it with a PlayStation 5 controller. When the custom application connected to DJI servers, approximately 7,000 robot vacuums across 24 countries simultaneously responded.

The vulnerability in DJI Home application backend authentication allowed unauthorized access to real-time camera feeds, microphone audio, 2D room maps, and device location data for affected ROMO robots. DJI reported discovering the authentication issue independently during routine internal security review in late January and received reports of the same vulnerability from two independent security researchers through its bug bounty program.

Reported impact

The vulnerability exposed the ability to access camera feeds, listen to audio through device microphones, and obtain complete 2D room maps and device locations for affected robot vacuums. DJI stated no evidence of user data misuse was found.

Index analysis

The security flaw was located in backend verification of the DJI Home application affecting ROMO robots and some DJI power bank products. The issue was identified through reverse engineering facilitated by AI coding tools and through the company's bug bounty program.

Governance controls

DJI completed patches and deployed updates to fully resolve the vulnerability. The company stated it has a bug bounty program with over 300 security researchers submitting reports, and plans independent third-party security audits and certification for ROMO and DJI Home App similar to processes used for drone products.

Classification

Organizations
DJI (subject), Anthropic
Category
Privacy and surveillance
Industry
Consumer products
Country
China
Harm status
Potential harm
Intent
Unintentional
Timing
Post-deployment
Harmed
Consumers
Consequences
None recorded
AI systems
Claude Code

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

每日经济新闻
大疆回应ROMO扫地机器人安全漏洞:已完成修复 2026-03-09 18:55
9 March 2026
Sep 2026Anthropic tests Claude Money bank account access featureWell documentedPrivacy and surveillanceThe IndependentMemeburnWindows Report | Error-free Tech Life+2Sep 2026Anthropic builds predictive surveillance system to track activists and protestsReportedPrivacy and surveillanceCommon DreamsFuturismAug 2026Private ChatGPT Conversations Used as Evidence in US Court CasesConfirmedPrivacy and surveillanceWashington PostInternational Business Times UKInternational Business TimesJul 2026Claude conversations and artifacts indexed in Google search resultsReportedPrivacy and surveillanceaxios.combbc.comJul 2026Claude shared chats indexed by Google search without explicit user consentAllegedPrivacy and surveillanceBeInCryptoInternational Business Times UKIT Security News - cybersecurity, infosecurity news+2