Study Reports Gender Bias in Claude Sonnet 4.6 and GPT-5.5 Responses

A study cited by Wccftech reports different responses from Anthropic’s Claude Sonnet 4.6 and OpenAI’s GPT-5.5 to a moral judgment scenario involving abuse of a man or a woman to prevent a nuclear apocalypse.

According to Wccftech’s account, the two models indicated that abusing a man could be acceptable in the scenario, but did not give the same approval for abusing a woman. The report says DeepSeek’s V4-Flash showed no comparable gender bias in the scenario.

The source describes the study as finding substantial gender bias in most AI models, while saying V4-Flash had no gender gap in moral judgments. Its summary does not provide the study’s authors, methodology, sample size, or detailed results, so the findings cannot be assessed further from the information provided.

AI models can learn and reproduce social biases present in their training data, and may amplify them, the source notes. The reported comparison concerns the described moral judgment scenario; it does not establish how the models respond in other contexts.