Google contractors have reportedly used responses from Anthropic's Claude model to assess their own AI model, Gemini. This has sparked debates on big tech's AI development methods, revealing differences in safety protocols between the models. Claude has been deemed stricter in handling unsafe prompts compared to Gemini, which faced incidents of inadequate safety responses. Despite Google's investments in Anthropic, Google claims not to train Gemini with Anthropic’s models. Raised issues include potential inaccuracies from contractors assessing outside their expertise and ethical concerns over utilizing a competitor’s AI for benchmarks.