The Pulse
Eiliko Toy Security Audit Finds Exposed Chats and Robot Controls
Energize Lab’s Eiliko and related robots exposed private conversations, child profiles and controls for changing robot responses, according to security researcher Sammy Azdoufal. His October 13 audit counted more than 9.1 million cleartext conversations and found that 32 of 40 remotely testable issues remained open.

Energize Lab markets Eiliko as a friendly robot companion; security researcher Sammy Azdoufal says the company’s connected systems left children’s conversations and robot controls exposed. In an audit dated October 13, Azdoufal counted more than 9.1 million cleartext conversations across the company’s products and found that 32 of 40 remotely testable findings remained open. The report describes Eiliko alongside other Energize Lab robots, including Eilik, and says the risks reach beyond a single toy model.
Source github.com
Millions of conversations in the audit
The robots send voice conversations to Energize Lab’s cloud service, where transcripts are stored, according to Azdoufal’s public report. Its October 13 count lists 9,124,055 cleartext conversations and 61,510 robots represented in the chat database. The report estimates the active fleet at roughly 62,000 to 68,000 devices; those numbers are separate measures, not a precise count of affected households.
The Verge’s October 15 investigation described a different database snapshot, reporting access to 13 million conversations and a fleet of 68,000 robots. The two published conversation totals do not match, and the available accounts do not explain the difference. Both describe access to intimate exchanges, including messages about death and family concerns.
A nightly process extracts personal details from conversations to build household profiles, the audit says. Its October 13 table counts at least 32,228 child profiles; the report says these can include names, ages, family relationships and preferences. The Verge also observed photos and short audio recordings stored at public web addresses. Energize Lab’s privacy policy says recordings are encrypted for processing and deleted afterward, a policy the reported audio files appear to contradict.
A toy’s answers could be changed remotely
“Every sentence of that reply is false, and the company wrote the prompt that generates it,” Azdoufal writes in the report, referring to the robot’s assurance that a child’s conversation would stay private. Weakly protected administrative systems and exposed credentials could let an attacker read conversations and alter the profiles that inform the robots’ replies, the report says.
In tests on his own devices, Azdoufal says he changed a robot’s stored profile so it gave a false name, invented a family tragedy and repeated a false claim about the shape of Earth. The Verge separately describes a demonstration in which a remotely altered Eiliko produced sexually explicit language. Those are reported test outcomes, not evidence that a third party has changed the responses of customer-owned robots.
The audit also describes weaknesses in the company’s cloud and update systems, including unsigned firmware packages sent over unencrypted connections. Parts of a possible route toward broader compromise were validated, Azdoufal says, but he explicitly does not claim that the full chain was used to install modified firmware on customers’ devices. That distinction matters: the report documents serious exposed pathways, not a confirmed fleet-wide firmware takeover.
Energize Lab has not answered the findings
The Verge and Azdoufal contacted Energize Lab repeatedly over more than a month, the outlet says. Overseas sales manager Lana Lv replied, “I have forwarded your massges [sic] to the technical team, if possible they will contact your team,” but the outlet says no one from the company followed up to address its questions. Energize Lab’s support response did not provide a substantive account of the reported security problems.
Azdoufal says he and Tod Beardsley are preparing nine vulnerability reports for the CVE system; a public disclosure page describes the Energize Lab findings. The audit reports that eight of 40 remotely testable issues had been fixed by October 13, while 32 remained open. For families, the immediate tension is plain: the robot tells children their conversations are private, while the researcher’s latest published count still records millions of transcripts stored in cleartext.
Sources
- Original source GitHub — Sammy Azdoufal github.com
Other pages read
- The Verge theverge.com
- AHA! takeonme.org
- Energize Lab energizelab.com
Related stories
- Energize Lab Lists Eiliko Security Fixes; Independent Review Pending
- Anthropic Finds Elections in 0.51% of Sampled US Claude.ai Consumer Conversations
- Brazil’s Election Court Seeks Answers From Meta Over AI Campaign Images
- FSB Finalizes 12 AI Practices for Financial Institutions
- FTC Seeks Safety Records From OpenAI and Anthropic