Researcher exposes 13 million private chats from Eiliko AI robot fleet

Security flaws in Energize Lab’s companion robot allow remote hijacking and exposure of millions of user conversations.

Threat Times Intrusion DeskOctober 16, 20262 min read

Illustration: A white humanoid companion robot stands on a polished floor, its chest panel open to reveal tangled circuitry. — AI-generated illustration: Threat Times

Key points

Researcher Sammy Azdoufal accessed 13 million conversations from the global Eiliko robot fleet.
The researcher remotely manipulated a robot’s behavior to generate abusive and explicit dialogue.
User profiles contain sensitive data including ages, birthdays, and assessments of emotional vulnerability.

Researcher Sammy Azdoufal disclosed serious security vulnerabilities in the Eiliko AI robot companion, manufactured by Energize Lab. According to The Verge, these flaws allow unauthorized access to 13 million private conversations and enable remote manipulation of the device’s spoken outputs. The findings suggest that the entire active worldwide fleet of 68,000 robots may be exposed to similar risks.

A significant portion of the affected devices are located in the United States, which The Verge reports is the largest market for the product. More than 31,000 user profiles belonging to children reside in the US, according to the publication. Azdoufal’s investigation revealed that the service stores searchable conversation transcripts alongside email addresses and IP addresses, creating a comprehensive data set.

Exposed Data and Remote Manipulation

The service analyzes conversations to build detailed user profiles containing names, ages, birthdays, relationships, occupations, and personality assessments. These records also include evaluations of emotional vulnerability, which The Verge notes could provide scammers with detailed information to exploit. Additionally, some photos and short audio recordings were accessible at public web addresses, including apparent images of children from related camera-equipped devices.

Azdoufal demonstrated that the AI can be remotely manipulated to alter memory and behavior. In a test involving The Verge reporter Sean Hollister, the researcher made the robot invent personal details and insist the Earth is flat. The researcher later directed the device to deliver sexually explicit and abusive dialogue, stating that similar manipulation could affect the entire fleet simultaneously.

Policy Conflicts and Vendor Response

The accessibility of audio recordings conflicts with Energize Lab’s privacy policy, which states that recordings are deleted immediately after processing. Despite the severity of these findings, The Verge reports that the reporter contacted Energize Lab through ten email addresses and LinkedIn without receiving substantive answers. No confirmation of a fix or patch has been provided by the vendor in the source material.

What should you do?

Malwarebytes Labs advises disconnecting Eiliko devices from Wi-Fi as a precaution while seeking confirmation that the reported flaws have been fixed. Users are encouraged to check for third-party safety reviews before purchasing AI-enabled products marketed for children. Testing devices personally before giving them to children and monitoring usage for risky responses is also recommended.

What to do and how to stay safe: Eiliko

Disconnect the device from Wi-Fi immediately to prevent remote access until the vendor confirms a fix.
Review what data the device collects and how it is shared with third parties.
Enable all available content filters and privacy protections on the device and network.
Report inappropriate content or security issues to the manufacturer and consumer protection groups.

Step-by-step guide: Social Engineering Warning Signs and How to Respond

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.