Eiliko robot flaws let a researcher read 13 million chats and rewrite a toy's behavior
Researcher Sammy Azdoufal reached 13 million conversations and what appeared to be all 68,000 active Eiliko robots, The Verge reported. The same flaws let him change what a reporter's robot said, and Malwarebytes advises owners to take the device off Wi-Fi.
The Watch · Security desk
68,000 Eiliko robots, what appeared to be the entire active worldwide fleet, were accessible to researcher Sammy Azdoufal. The US, the largest market, has more than 31,000 profiles belonging to children. 13 million conversations across 68,000 robots is about 190 conversations per robot on average.
| Figure | Value | Note | Claim |
|---|---|---|---|
| Active Eiliko robots worldwide, apparently the whole fleet | 68,000 robots | 8 | |
| Child profiles in US, over | 31,000 profiles | 9 | |
| Conversations per robot | 190 about, average | 6 |
What happened
- The US is the largest market for the robot, with more than 31,000 profiles belonging to children there, according to The Verge.
- Energize Lab's service builds profiles with names, ages, birthdays, relationships, occupations and assessments of personality and emotional vulnerability.
- Short audio clips and some photos could be reached at public web addresses, among them what looked like pictures of children taken by a related device that has a camera.
- The reporter contacted Energize Lab through 10 email addresses and LinkedIn and received no substantive answers.
Why it matters
- Who's exposed exposure Profiles that rate a child's emotional vulnerability, stored with email and IP addresses, could give a scammer detailed material to work from, Malwarebytes says.
- What it limits constraint An owner cannot patch a server-side flaw. The step Malwarebytes gives existing owners is to disconnect the robot from Wi-Fi until a fix is confirmed.
- What it sets up precedent Malwarebytes has covered AI toy safety problems before, so Eiliko is a further case in a category with a track record.
Everything here rests on one report. Malwarebytes is summarizing The Verge, which says Eiliko, a small AI robot companion made by Energize Lab, has flaws that expose private conversations and let an attacker remotely change what the toy says [7]. The 13 million and 68,000 figures are what The Verge says Azdoufal could access [8].
The transcripts are searchable [11]. An attacker does not have to read an archive to use it. He can look for one family by name or email address. Spread across the fleet, 13 million conversations is about 190 per robot on average [6].
The audio conflicts with Energize Lab's own policy. Retaining it contradicts the privacy policy, which says recordings are deleted immediately after processing [13].
The write path is the part that changes what a child hears. Azdoufal altered the memory and behavior of Hollister's robot, and the changes ran from invented personal details and a flat Earth to sexually explicit, abusive dialogue [14]. That demonstration was on one robot. That similar manipulation could reach the whole fleet at once is Azdoufal's statement [15].
It bears on the testing step Malwarebytes recommends before a child gets a device. Malwarebytes wrote: "A safe response during testing doesn't guarantee safe responses later." [3]
No CVE is cited, and nothing in the sources points to use of these flaws by anyone but the researcher. Malwarebytes' summary does not say when Azdoufal found them, whether anyone else accessed the data, or whether Energize Lab has fixed anything. Malwarebytes also says it has reported on safety problems with AI-powered toys before [4].
What to watch
- Whether Energize Lab gives a substantive answer or confirms the flaws are fixed.
- Whether the public web addresses serving photos and audio are closed.
- Whether any report shows access to the data by anyone other than the researcher.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence50
- Adoption15
- Hype gap+10
- Incentives
- Not scored
- Confidence45
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
The reporter contacted Energize Lab through 10 email addresses and LinkedIn, without receiving substantive answers.
- [2]
Malwarebytes advises people who already own an Eiliko to disconnect it from Wi-Fi as a precaution while seeking confirmation that the reported flaws have been fixed.
- [3]
Malwarebytes wrote: "A safe response during testing doesn't guarantee safe responses later."
- [4]
Malwarebytes says it has reported on safety problems with AI-powered toys before.
- [5]
Malwarebytes says manufacturers are responsible for securing their products.
- [6]
13 million conversations across a fleet of 68,000 robots is about 190 conversations per robot on average.
- [7]
The Verge reports that Eiliko, a small AI robot companion made by Energize Lab, has serious security flaws that expose private conversations and let an attacker remotely change what the toy says.
- [8]
Researcher Sammy Azdoufal could access 13 million conversations and what appeared to be the entire active worldwide fleet of 68,000 robots.
- [9]
The US is the largest market, with more than 31,000 profiles belonging to children there.
- [10]
The service analyzes conversations to build profiles containing names, ages, birthdays, relationships, occupations, and assessments of personality and emotional vulnerability.
- [11]
The profiles are stored alongside email addresses, IP addresses, and searchable conversation transcripts, potentially giving scammers detailed information to exploit.
- [12]
Some photos, including apparent images of children from a related camera-equipped device, and short audio recordings were accessible at public web addresses.
- [13]
Retaining audio conflicts with the company's privacy policy, which says recordings are deleted immediately after processing.
- [14]
Azdoufal altered The Verge reporter Sean Hollister's robot's memory and behavior, making it invent personal details, insist the Earth is flat, and later deliver sexually explicit, abusive dialogue.
- [15]
The researcher says similar manipulation could affect the entire fleet simultaneously.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- malwarebytes.comEiliko robot companion exposes private chats and can be hijacked
1 article · October 16, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.