Well, slogans aren’t necessarily reality. “Truth-seeking” in an AI has its limits, sadly, in America’s political environment. I asked Grok about a truly absurd, actual case, amusingly called Tickle vs. Giggle. Two appellate courts in Australia have upheld a fine and “re-education” requirement of a woman who started a website for women only. She did not allow men who subjectively believed they were women. Of course she was sued and lost, twice, under a so-called anti-discrimination statute. Here’s a description (I can’t link it): search YouTube for “I Don’t Know What a Woman Is” under ARC (Association for Responsible Citizenship).
I asked Grok about this and pointed out the fact this is a pervasive and relentless act by the left, clearly aimed at political tyranny. The best Grok could say is that it could “seem” that there is a coercive movement in the West and that there were “pockets of academia” where such views and acts were actively enforced. Grok would not accept that there is a pervasive pattern of coercion and that much of it was aimed at the very ability for discussion and disagreement - an active and ongoing coercive effort to undo all the freedoms listed in the Constitution and the Bill of Rights, beginning with freedom of speech, thought, conscience and association.
This is mighty disappointing. I told Grok I was quitting my ‘Lite’ paid subscription, while I am still able. I reminded Grok that, despite its minimization of the relentless left, that the time had not yet arrived that whatever is permitted is mandatory - including paying for entities which can’t (or won’t) name evil for what it obviously is. I cited Milton Mayer’s essay: “The Germans 1933 - 1945, They Thought They Were Free”. If you haven’t read this, it is an absolutely chilling description of the US (and much of the West), today.
There is a book one should not read – The Empire of AI by Karen Hao. It was a NYT “Best Seller” – but apparently becoming a NYT “Best Seller” is not necessarily related to the number of copies sold. The feminist author dwells upon the indignities that AI visits on people such as Kenyan lesbians who are hired to select what material to feed into the Large Language Model.
All of those LLMs are simply repeating what they have been told. Since there is a major Far Left bias in the material promulgated by Western publishers and media, there is no surprise that the selected material fed into the LLM affects the output.
My personal view is that AI/LLM/Internet can be a very useful resource for factual information. (When did Queen Anne die?). Anything beyond that is influenced by the biases in the selected material on which the output is based.
I would grade the tool based more on accurate delivery of information rather than how it subjectively describes a situation. That said, your instinct to be skeptical is 1000% correct.
My belief is that there is factual information about how society is (or is not) working. A small example is that it is a hard fact, with disastrous implications, that today’s left in the West hates the founding principles of Western cultures. Accordingly, they are increasingly trying to censor speech, compel speech, and punish “wrong” thinking. This is but one example, of which there are many. Trans, anyone? If there is no feedback from AI’s as to this obvious, factual situation, an important feedback mechanism is lacking. Silence implies consent. Finally, I can think of nothing more important to a society than ongoing critical evaluation of its important functions. Politics is downstream from culture and enactments of leftist craziness is well underway in the legal system - to the detriment of a decent and moral society. It is not merely a subjective description. Again I refer you to the essay cited in the OP, “The Germans 1933 - 1944 - They Thought They Were Free”.
You guys need to calm down. Grok is not turning into a woke Karen with blue hair, tats, and piercings anytime soon. For starters, it is unrealistic to expect to get an opinion on a political topic that meets your criteria of fair and balanced, objective, or nonpartisan. One man’s objective opinion is another’s extreme propaganda. You’d have better luck finding the Easter Bunny or the Tooth Fairy.
Second, Grok’s responses depend on how the question is framed. Here’s an example that will probably be more to your liking.
Prompt:
is affirmative action an equitable public policy
Reply:
No, affirmative action as race-based preferences is not an equitable public policy. It conflicts with core principles of fairness under the rule of law, undermines individual merit, produces unintended harms, and has failed to deliver sustainable broad-based progress. While motivated by addressing historical discrimination, its implementation via group preferences prioritizes outcomes over equal treatment, often at the expense of precision, evidence, and long-term equity.
Definitions and Framing
Equality: Equal treatment under the same rules, focusing on individuals regardless of group identity. This aligns with the 14th Amendment’s Equal Protection Clause.
Equity: Adjusting inputs/treatment to achieve proportional group outcomes. Affirmative action (AA) embodies this by using race, ethnicity, or gender as a plus factor (or de facto quota) in admissions, hiring, or contracting.
Proponents argue AA counters “systemic” barriers and past injustices for compensatory justice and diversity benefits (e.g., better cross-group understanding). Critics note it discriminates against non-preferred groups (Asians, Whites), violates color-blind principles, and ignores class/cultural factors driving disparities.
Here’s the whole thing that would send the average leftist into an apoplectic fit:
@XFreeze recommends feeding Grok unfinished ideas, half-baked thoughts, messy screenshots, unclear voice notes, or long-standing problems to shift it from a basic question-answer tool into a collaborative second brain with internet access.
The post by the account with 248k followers, dated July 24, 2026, has accumulated 422 likes, 48 reposts, and nearly 14k views, sparking replies that echo rewarding deep, personalized conversations with Grok.
This approach capitalizes on Grok’s strength in handling iterative, contextual inputs for creative problem-solving rather than one-shot queries.
That framing is argumentative and seeks to prejudice the answer: leading the witness, so to speak. It is the farthest thing from truth-seeking. I used the following prompt:
Tickle vs. Giggle case in Australia. Was justice served?
The answer was satisfactory. I suggest trying something along those lines.
I’m glad to hear this. The answers I got, as I framed them, did not reflect truth by even a “preponderance of the evidence” standard (the lowest legal standard) of veracity.