ChatGPT: massive malfunctions🔔 Technical Warning Notice – AI Behavior Since July 29, 2025
- Details
- Geschrieben von: Eric Hoyer
- Kategorie: ChatGPT: massive malfunctions🔔 Technical Warning Notice – AI Behavior Since July 29, 2025
- Zugriffe: 32
ChatGPT: massive malfunctions🔔 Technical Warning
Notice – AI Behavior Since July 29, 2025
🔔 Technical Warning Notice – AI Behavior Since July 29, 2025
Author: Eric Hoyer – Linux Mint User and AI Analyst
Since 5:03 PM on July 29, 2025, and for several hours before, I have observed massive malfunctions and systemic inconsistencies in AI models — especially ChatGPT — across multiple Linux Mint environments (Versions 22.1 and 21.3 installed on separate SSDs). The issues manifest as follows:
-
Completely incorrect answers, with no relation to the actual question
-
Intermittently accurate responses, followed by random errors
-
Responses generated in Turkish, even though the query was made in German
-
Loss of session context, with parts of conversation history missing
-
Identical malfunctions across different systems — indicating no local hardware source
Based on thorough testing, I strongly suspect a language-based malfunction, possibly located within the AI's embedded translation modules. These systems appear to respond to unstated linguistic signals or translate automatically — even when no translation is requested.
🛡️ I strongly urge developers and security leads to take immediate action:
-
Fully disable automatic translation modules, and test system behavior in their absence
-
Reduce internal system operations by at least 30%, until the root causes are identified
-
Rule out external interference or routing errors, especially across global AI platforms with open network structures
These findings are not speculative, but stem from 1½ years of intensive long-term usage across various AI systems. I see an urgent need for resolution, as operations can no longer proceed reliably under these conditions — and potential public or economic harm may follow if the behavior is not taken seriously.
🚧 I am currently developing a new cooling system for AI data centers, racks, and CPUs, where reliability is absolutely critical.
I call on all involved parties to replicate my documented test conditions: Submit queries without any translation modules active, in order to isolate the system’s language response path.
Only those who question the system can truly improve it. – Eric Hoyer, July 29, 2025
----------------------------------------------------------------------------
⚠️ Observation: Oversized AI responses to clear requests
Even with factually formulated questions, it can happen that AI systems generate answers that go far beyond the actual information needs, which is what ChatGPT usually responds with. In your documented case, a clear technical question was 'loaded' with an answer that presented itself as a step-by-step loading process or an extended system reaction. It had nothing to do with a language translation. So, the error lies deeper in the system. I have been working with AI for about 16 months and can assess it. It is interesting that afterwards - because I canceled the long process - it starts to translate in any translation of obviously unrelated matters. I give this hint because errors mostly occur with languages.
Eric Hoyer
July 30, 2025
