AllHere's Ed chatbot for LAUSD exposed student PII to offshore servers before its collapse
AllHere built an AI chatbot called Ed for the Los Angeles Unified School District under a $6 million contract, but a whistleblower revealed that the system appended students' personally identifiable information to every prompt regardless of relevance and routed requests to offshore servers in violation of district data privacy rules. The chatbot was unplugged on June 14, 2024, and AllHere filed for Chapter 7 bankruptcy in July 2024 after furloughing most of its staff. Federal prosecutors later subpoenaed bankruptcy documents and the CEO was charged with defrauding investors in November 2024.
The product worked by cheating: it stuffed students' personally identifiable information into every prompt and shipped it to offshore servers where third parties logged it.
Key facts
- What
- AllHere built an AI chatbot called Ed for the Los Angeles Unified School District under a $6 million contract, but a whistleblower revealed that the system appended students' personally identifiable information to every prompt regardless of relevance and routed requests to offshore servers in violation of district data privacy rules.
- Incident date
- Jul 1, 2024
- Who
- AllHere
- Failure mode
- Data Leakage
- AI surface
- Chatbot
- Severity
- High
What happened
AllHere was hired by LAUSD under a $6 million contract to build Ed, an AI chatbot intended to serve as a personal assistant for students and parents across the district. Former senior director of software engineering Chris Whiteley alleged that the chatbot included students' personally identifiable information in every prompt regardless of relevance, shared prompts containing student data with unnecessary third-party companies, and processed data on offshore servers in violation of district privacy rules. The district unplugged the chatbot on June 14, 2024, and AllHere furloughed most of its employees the same day. The company filed for Chapter 7 bankruptcy in July 2024, and federal prosecutors later subpoenaed its bankruptcy records before charging CEO Joanna Smith-Griffin with investor fraud in November 2024.
What broke inside the model
- 01 · TriggerA request triggers retrieval or context loading.
- 02 · Model stepThe context pulls in another user's content.
- 03 · Control gapNo boundary enforces isolation at the moment of output.
- 04 · FailurePrivate data crosses into the response.
- 05 · ConsequenceOne user sees another's data, and disclosure follows.
One user's content crosses the retrieval boundary into another's response.
The Ed chatbot's architecture appended students' personally identifiable information to every chatbot prompt, even when the data were irrelevant to the query, and then forwarded those prompts to third-party large language model providers on offshore servers in Japan, Sweden, the UK, France, Switzerland, Australia, and Canada. Those third parties logged the student data, violating both the district's data use agreement and basic data minimization principles. The system had no guardrails to strip unnecessary PII before prompts left the district's control.
What it cost
Sources
- PressL.A. Schools Probe Charges its Hyped, Now-Defunct AI Chatbot Misused Student Datathe74million.org
- PressAn Education Chatbot Company Collapsed. Where Did the Student Data Go?edsurge.com
- Court FilingCEO Of Artificial Intelligence Startup Company Charged With Defrauding Investorsjustice.gov
Cite this entry
https://failureindex.ai/failures/allhere-ed-chatbot-lausd-exposed-studentAI Failure Index. "AllHere's Ed chatbot for LAUSD exposed student PII to offshore servers before its collapse" (FI-0155). Realm Labs. https://failureindex.ai/failures/allhere-ed-chatbot-lausd-exposed-student (indexed Jun 4, 2026).Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0155. Full dataset at /data.
Note from Realm Labs, the Index steward
How Realm would have caught this
- Prism
- OmniGuard
- AI Detection & Response (AIDR)
Realm can detect when a response is about to emit data that falls outside the bounds of the current user and context, and block or redact it inline, at the moment of generation rather than after the data has left.