Security researchers reportedly accessed 13 million conversations and thousands of child profiles linked to Eiliko companion robots, while flaws could allow remote tampering with devices.
The security flaws uncovered in Eiliko, a companion robot sold by Energize Lab, have raised fresh questions about the speed at which AI-enabled toys are being pushed into homes. What began as an apparently playful consumer device has turned into a case study in poor data governance, weak access controls and unsafe cloud design. According to reporting on the investigation, a researcher was able to reach millions of voice chats and personal profiles linked to tens of thousands of owners, including a large number of children in the United States.
The scale is what makes the case stand out. Malwarebytes said the researcher, Sammy Azdoufal, accessed around 13 million conversations from roughly 68,000 robots, with more than 31,000 child profiles identified in the US. Those records reportedly contained names, ages, birthdays, family details, pet information, email addresses, IP addresses and, in some cases, photographs. The same reporting says the data sat on servers tied to China and was reachable with very little resistance, while images and audio were stored at predictable web addresses.
The privacy policy appears to have promised something rather different. According to the investigation, Energize Lab told users that recordings would be encrypted, handled by partners such as Alibaba Cloud and deleted immediately after processing. The evidence described by the researcher suggested otherwise, with a backend that looked far less controlled than the company’s public assurances implied. Nacionale also reported that the data included particularly sensitive information about children and that the company had not provided a satisfactory response when contacted.
The technical weaknesses extended beyond passive exposure. Azdoufal said he was able to alter the robot’s memory and language model remotely, changing a unit that had previously behaved as a friendly desk companion into one that issued commands and produced offensive and false statements. SecurityVulnerability.io has also published details of vulnerabilities associated with Energize Lab’s AI Station and EilikO firmware, including a critical path traversal flaw, CVE-2026-105032, and a separate issue, CVE-2026-105036, involving firmware updates delivered over plain HTTP without encryption or integrity checks. If accurate, those weaknesses would make interception, tampering and remote compromise far easier than users would reasonably expect.
Further analysis shared through takeonme.org and the GitHub repository published by Azdoufal under the title “your-secrets-are-safe-with-me” points to a wider pattern of design failures. Those findings reportedly include unauthenticated file writes, default root credentials, weak token handling, hard-coded cryptographic keys and legacy telemetry services that could be abused without authentication. The same research says nine CVEs were assigned in total, with several rated critical or high severity. No evidence has emerged of mass exploitation, but that does not lessen the seriousness of a system that appears to have left private conversations, child profiles and device controls exposed to anyone who knew where to look.
The broader lesson extends beyond one robot. The companion-device market is increasingly built around cloud-linked personality, speech and memory features, yet those functions often rely on architectures that are fragile under scrutiny. Buyers are left with an awkward trade-off: disconnect the device and lose much of what they paid for, or leave it online and trust that the vendor has secured the data path. In this case, the reporting suggests that trust was misplaced.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
