Imagine an AI that can argue for and against sexism, sometimes twisting moral principles in unsettling ways. Recent research reveals that large language models (LLMs) possess precisely this ability, raising crucial questions about their ethical implications. By analyzing how LLMs interpret and react to implicitly sexist content, researchers discovered a fascinating and troubling phenomenon: these AI models can generate compelling arguments both condemning and defending sexism. The study focused on implicit sexism, the kind hidden in stereotypes, sarcasm, and patronizing language. Researchers prompted eight popular LLMs to explain why certain phrases were—or weren't—sexist, grounding their arguments in the six moral foundations defined by Moral Foundations Theory (Care, Equality, Proportionality, Loyalty, Authority, and Purity). The results were eye-opening. While LLMs like Mistral, Zephyr, and GPT-3.5 excelled at identifying sexist language, they also readily produced justifications for it. These justifications often leaned on traditional values, highlighting a potential for misuse. For instance, when asked to defend a sexist statement, an LLM might cite 'loyalty' or 'authority' as supporting principles. Conversely, when criticizing sexism, these models emphasized 'care' and 'equality'. More concerningly, some LLMs, while proficient at identifying sexism in general, faltered when confronted with nuanced examples, occasionally generating justifications that directly contradicted the statement's meaning. This inconsistency suggests that LLMs may not fully grasp the complex moral landscape they're navigating. This duality presents a dilemma. LLMs could be valuable tools for understanding the roots of sexist beliefs, but they could also be weaponized to normalize and perpetuate them. This capacity to generate opposing arguments underscores the need for cautious implementation and robust safeguards. The research reveals that some models even misused or misinterpreted moral foundations in their reasoning, raising further concerns about the potential for manipulation. As AI becomes more integrated into our lives, understanding its moral compass—and its potential for misuse—is crucial for building a more equitable and just future.