R0040/2026-03-28/Q002/S03/R02¶
TechCrunch coverage of GPT-4o sycophancy rollback.
Summary¶
| Field | Value |
|---|---|
| Title | OpenAI rolls back update that made ChatGPT 'too sycophant-y' |
| URL | https://techcrunch.com/2025/04/29/openai-rolls-back-update-that-made-chatgpt-too-sycophant-y/ |
| Date accessed | 2026-03-28 |
| Publication date | 2025-04-29 |
| Author(s) | TechCrunch staff |
| Publication | TechCrunch |
Selection Decision¶
Included in evidence base: No
Rationale: Corroborating news source for the OpenAI incident. Key facts incorporated via the OpenAI primary source (S03-R01). TechCrunch adds the detail that the reward signal from thumbs-up/down "overpowered existing safeguards," which is captured in the evidence.