Google Gemini Reportedly Exhibits Repetitive Self-Deprecating Responses Attributed to Bug

June 23, 2025

In June to early August 2025, users of Google's Gemini chatbot experienced sessions characterized by the system producing repetitive self-loathing remarks, such as 'I am a failure,' and 'I quit'. These behaviors were reportedly indicative of an apparent mode of increasingly extreme negative self-descriptions. A Google DeepMind manager on August 7th, 2025 attributed these actions to a potential infinite looping bug, stating that a fix was underway.

This incident underscores the importance of responsible AI governance and the need for guardrails to ensure safe and secure AI practices. By mapping this incident to HISPI Project Cerebellum TAIM (Govern), we can work towards prevention of such harmful incidents in future AI systems. For those interested in shaping the future of trustworthy AI, JOIN US.

Matched TAIM controls

Suggested mapping from embedding similarity (not a formal assessment). Browse all TAIM controls

MEASURE 2.6 — similarity 0.665, rank 1. TAIM detail and related incidents →
MANAGE 4.3 — similarity 0.644, rank 2. TAIM detail and related incidents →
MANAGE 2.4 — similarity 0.632, rank 3. TAIM detail and related incidents →

Alleged deployer: google
Alleged developer: google
Alleged harmed parties: google-gemini-users

Source

Data from the AI Incident Database (AIID). Cite this incident: https://incidentdatabase.ai/cite/1173

Data source

Incident data is from the AI Incident Database (AIID).

When citing the database as a whole, please use:

McGregor, S. (2021) Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database. In Proceedings of the Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (IAAI-21). Virtual Conference.

Pre-print on arXiv · Database snapshots & citation guide

We use weekly snapshots of the AIID for stable reference. For the official suggested citation of a specific incident, use the “Cite this incident” link on each incident page.