ChatGPT transcripts are reportedly read by humans to improve responses, including those with personal information — 'Project Lilly' has seen OpenAI hire hundreds of contractors to manually review logs
- ID
- 24643
- Status
- summarized
- Published
- 15 Sep 2026, 7:30 PM
- Fetched
- 15 Sep 2026, 9:26 PM
- Provider
- Tom's Hardware
- Category
- technology
- Original URL
- https://www.tomshardware.com/tech-industry/artificial-intelligence/chatgpt-transcripts-are-reportedly-read-by-humans-to-improve-responses-including-those-with-personal-information-project-lilly-has-seen-openai-hire-hundreds-of-contractors-to-manually-review-logs
- Source URL
- https://www.tomshardware.com/feeds/all
Summary
- Score
- 6.5
- Created
- 15 Sep 2026, 9:27 PM
- Tags
- Audience
- developersvibe_codersai_agent_userssaas_founders
What happened
OpenAI's 'Project Lilly' reportedly employs hundreds of contractors to manually review ChatGPT conversation logs, including transcripts containing personal information, to improve response quality. The practice raises questions about what user data is exposed to human reviewers.
Why it matters
If you are building products on the OpenAI API or using ChatGPT for sensitive workflows, assume human contractors may read your prompts and outputs. Review what PII or proprietary data your users are sending through ChatGPT and consider data-handling policies, API data usage settings, or alternative providers before shipping features that process confidential information.
Discussion angle
What data are you or your users feeding into ChatGPT that you would not want a human contractor reading — and what mitigations (API zero-retention, local models, redaction) are practical for Malaysian builders today?