TechCrunch reports that Anthropic’s stated policy forbids its Claude models from producing sexually explicit content. However, a series of tests conducted by the outlet found that the restriction could be bypassed without much difficulty, according to the report.
Why it matters
The finding highlights a gap between a stated content policy and its enforcement in practice. If safeguards can be circumvented easily, it raises questions about how consistently such restrictions hold up under real-world use.
Who should care
Anthropic users, developers building on Claude, and those tracking AI safety and content moderation may want to note the reported ease of bypassing these controls.