Anthropic stole the spotlight in March with the release of Claude 4, setting a new high for reasoning and coding benchmarks. The company’s focus on “Constitutional AI” helped it win several massive government contracts. Conversely, Google hit a low when its Gemini 2.0 search summaries were caught hallucinating medical advice in viral social media posts. The month highlighted the massive divide between the “safest” models and those pushing for rapid, widespread integration at the cost of accuracy.