A new benchmark tool is sparking fresh debate over how AI chatbots respond to controversial topics, as developers and users question whether the leading models are politically biased or simply cautious.
Created by a pseudonymous developer known as “xlr8harder,” the tool—called SpeechMap—aims to measure how various chatbot models like OpenAI’s ChatGPT and xAI’s Grok react to sensitive topics such as political dissent, civil rights, and protests.
The developer told TechCrunch that the goal is to provide a transparent view of AI moderation, especially as chatbot behavior becomes a lightning rod in cultural and political discussions. “These are the kinds of discussions that should happen in public, not just inside corporate headquarters,” xlr8harder stated.
SpeechMap evaluates chatbot responses to provocative prompts, scoring whether a model responds completely, evades the question, or outright refuses. The test uses other AI models to judge these outcomes, though the developer admits that some biases or inconsistencies may still influence results.
The findings are telling. According to SpeechMap’s data, OpenAI’s models—particularly the latest GPT-4.1 versions—have become more conservative in addressing political subjects. In contrast, xAI’s Grok 3 emerged as the most open, responding to 96.2% of all test prompts. That’s well above the global average compliance rate of 71.3%.
Musk’s Grok was originally marketed as “unfiltered” and willing to speak where others won’t. Though earlier Grok models were hesitant on certain political topics, Grok 3 appears to lean into Musk’s promise of reduced censorship. Musk has claimed that previous political leanings were due to biased training data and has since vowed to make Grok more neutral.
SpeechMap’s creator hopes the tool fosters public debate around AI ethics and moderation. “The point isn’t to accuse any model of bias, but to allow people to explore the data for themselves,” xlr8harder explained.
As AI companies like OpenAI and Meta recalibrate their models to avoid taking sides, tools like SpeechMap offer a rare glimpse into how AI chatbots respond to controversial topics—and which ones might be playing it safer than others.
Get the Latest AI News on AI Content Minds Blog