Back to today’s briefing

Story perspectives

AI Hallucinations Soar: 79% False Info in Tests

5/9/2025

45 7

Summary

Story summary

Recent advancements in AI reasoning systems, like OpenAI's models, have sparked concerns over rising hallucination rates, with the o4-mini model generating false information 79% of the time in tests. While progress is evident, the struggle for accuracy continues, raising questions about the reliability of AI in critical situations.

1 of 1

Story summary

Recent advancements in AI reasoning systems, like OpenAI's models, have sparked concerns over rising hallucination rates, with the o4-mini model generating false information 79% of the time in tests. While progress is evident, the struggle for accuracy continues, raising questions about the reliability of AI in critical situations.