On April 24, 2026, AI company Anthropic released a technical review report, admitting that three adjustments at the product level led to a decline in the performance of the Claude model, but denied intentionally 'reducing intelligence' to save computing power. Currently, the related bugs have been fixed.
