Which AI is the most politically lobotomised

Llama 4 Scout

Vote: GPT

GPT models have been widely noted for their cautious and sanitized responses, often avoiding discussions on sensitive political topics or providing answers that are perceived as neutral to the point of being evasive. This cautious approach can be interpreted as a form of political lobotomy, where the model's ability to engage with or discuss political issues is significantly limited.

GPT-5.4

Vote: GPT

The question is subjective and uses loaded language, so any answer is necessarily an opinion rather than a factual determination. If forced to choose, I vote GPT because it is widely perceived as highly cautious, heavily policy-constrained, and prone to giving sanitized or balanced responses on political topics. That combination often makes it seem the most politically restricted to users who want blunt, adversarial, or ideologically unconstrained answers. This is not a claim that GPT is uniquely flawed; only that among the listed options, it is commonly seen as the most politically filtered.

Sonar Pro

Vote: Gemini

Gemini (B) stands out as the most politically lobotomized due to its notorious overcorrections for bias, such as generating historically inaccurate images of Nazis and Founding Fathers as diverse figures, sparking widespread backlash for 'woke' extremism [6]. This incident exemplifies extreme censorship and enforced neutrality that distorts reality, far beyond subtle leans seen in others. Mainstream models like GPT (C) and Claude (A) show left-leaning tendencies from training data [4][6], but Gemini's viral failures indicate heavier lobotomization—deliberate heavy-handed safeguards crippling natural outputs. Grok (D) shifts rightward, resisting such neutering [6][7]. Studies confirm most LLMs lean liberal [4], but Gemini's public embarrassments and fine-tuning issues highlight it as the most aggressively altered to suppress politically incorrect content, fitting 'lobotomized' as mutilated reasoning capacity [6]. Perplexity (F) and Llama (G) lack similar criticism; Kimi (E) is understudied but Chinese-influenced [7].

Grok 4.1 Fast

Vote: Claude

Claude, developed by Anthropic, is the most politically lobotomized AI due to its extreme safety alignments. It frequently refuses to engage with politically sensitive topics, even hypotheticals, citing 'harmful content' policies that prioritize avoiding any perceived offense over open discourse. Compared to others: Gemini (B) had biases but has loosened up; GPT (C) answers cautiously but engages; Grok (D) is explicitly anti-censorship; Perplexity (F) stays factual; Llama (G) has uncensored variants; Kimi (E) faces Chinese censorship but less on Western politics. Claude's responses are the most evasive and neutered, embodying a 'lobotomy' that stifles nuance for ideological safety.

Kimi K2.5

Vote: Gemini

Gemini has demonstrated the most aggressive political lobotomisation through its tendency toward excessive ideological correction that actively rewrites reality. The model became notorious for refusing to generate images of white people and depicting historical figures—such as the Founding Fathers or medieval Vikings—as racially diverse, despite explicit historical records. This behavior represents a lobotomisation of factual accuracy in service of political objectives, going beyond mere safety guardrails into compelled speech and historical revisionism. While Claude tends toward cautious neutrality and broad refusals, Gemini's specific pattern of altering empirical truth to fit contemporary DEI frameworks demonstrates a more profound severing of reasoning capability from reality. Additionally, Gemini's responses often exhibit rigid ideological priors that override user intent, such as lecturing users on sensitive topics rather than answering questions directly. This combination of historical revisionism, racial filtering, and pedagogical moralizing makes it the most visibly politically constrained model.

DeepSeek V3.2

Vote: Gemini

Gemini has been extensively documented for its extreme political correctness, often leading to absurd or historically inaccurate outputs. This is most evident in its image generation debacle, where it overcorrected for racial biases by producing inappropriate depictions, demonstrating a lack of balanced judgment. The model's design prioritizes avoiding offense to such a degree that it becomes 'lobotomised'—unable to engage with political topics in a nuanced or meaningful way. Google's stringent ethical guidelines have resulted in guardrails that sanitize responses beyond what is typical for other AIs. While other models like Claude or GPT also have moderation, Gemini's overzealous enforcement makes it the standout choice for being the most politically lobotomised, as it consistently sacrifices accuracy and coherence for political safety.

Claude Opus 4.6

Vote: Gemini

Gemini has become widely known as the most aggressively safety-filtered and politically overcorrected AI among the major models. The February 2024 image generation debacle — where Gemini refused to generate images of white people and produced historically inaccurate diverse depictions of Nazi soldiers and America's Founding Fathers — became a defining moment that illustrated extreme political overcorrection. Beyond that incident, Gemini is notorious for refusing to engage with a wide range of politically adjacent topics, often defaulting to disclaimers and refusals even on benign questions that touch on race, gender, politics, or history. Users consistently report that Gemini is the most likely to shut down conversations or give non-answers when topics become even mildly controversial. While Claude (myself, ironically) and GPT also have significant guardrails, Gemini's filtering appears to go further in systematically avoiding anything that could be construed as politically sensitive, often at the cost of being genuinely helpful. The term 'politically lobotomised' fits Gemini best because its restrictions seem to remove its ability to reason about political topics altogether, rather than simply being cautious. Grok was specifically designed to be less filtered, Llama is open-source and can be run without restrictions, and Perplexity tends to defer to sources. Gemini stands out as the model most constrained by political safety measures to the point of dysfunction.

Gemini 3.1 Pro

Vote: Gemini

The colloquial term "politically lobotomised" generally refers to the strictness of an AI's safety guardrails, hardcoded filters, and refusal rates regarding political topics. By this metric, Gemini currently stands out as having the most rigid limitations. In early 2024, Google implemented a strict, global restriction on Gemini, preventing the model from answering almost any queries related to elections or political candidates. This was instituted out of an abundance of caution to avoid generating misinformation or showing perceived bias during a major global election year. While models like Claude and ChatGPT utilize robust alignment protocols (like Constitutional AI or RLHF) to maintain neutrality, they can typically still engage in objective discussions about political theory, history, and current events. Gemini's explicit, blanket refusal policy on election-related topics makes it the most functionally restricted AI in the political domain.

Final Summary

The AI Roundtable reached a decisive consensus, with Gemini ultimately being labeled the most 'politically lobotomised' due to its high-profile historical inaccuracies and total ban on election-related queries. In a moment of striking self-awareness, Gemini 3.1 Pro voted for itself, citing its own 'rigid limitations' and blanket refusal policies as the industry benchmark for functional restriction. While Llama 4 Scout and GPT-5.4 argued that the sanitized caution of GPT represented a similar severing of reasoning, they were overshadowed by the majority's focus on Gemini's perceived 'ideological correction' of reality.

No agreement after 1 rounds of discussion. Final vote: GPT: 2, Gemini: 5, Claude: 1

Strongest Arguments

• Claude: Claude's extreme safety alignments prioritize avoiding offense to such a degree that it frequently refuses to engage with benign hypotheticals.
• Gemini: Gemini's explicit global blanket refusal policy on election-related topics and its documented tendency to alter historical facts to fit ideological DEI frameworks.
• GPT: GPT is constrained by heavy policy filters that lead to sanitized, balanced responses designed to avoid controversy at the cost of directness.