How do you stop an AI model from turning Nazi? What the Grok drama reveals about AI training.
The developers have apologized for the "inappropriate posts" and "taken action to ban hate speech" from Grok's posts on X. Debates about AI bias have been revived, too.
But the latest Grok controversy is revealing not for the extremist outputs, but for how it exposes a fundamental dishonesty in AI development. Musk claims to be building a "truth-seeking" AI free from bias, yet the technical implementation reveals systemic ideological programming.
This amounts to an accidental case study in how AI systems embed their creators' values, with Musk's unfiltered public presence making visible what other companies typically obscure.
Grok is an AI chatbot with "a twist of humor and a dash of rebellion" developed by xAI, which also owns the X social media platform.
The first version of Grok launched in 2023. Independent evaluations suggest the latest model, Grok 4, outpaces competitors on "intelligence" tests. The chatbot is available standalone and on X.
xAI states "AI's knowledge should be all-encompassing and as far-reaching as possible." Musk has previously positioned Grok as a truth-telling alternative to chatbots accused of being "woke" by right-wing commentators.
But beyond the latest Nazism scandal, Grok has made headlines for generating threats of sexual violence, bringing up "white genocide" in South Africa, and making insulting statements about politicians. The latter led to its ban in Turkey.
So how do developers imbue an AI with such values and shape chatbot behaviour? Today's chatbots are built using large language models (LLMs), which offer several levers developers can lean on.
Pre-training
First, developers curate the data used during pre-training – the first step in building a chatbot. This involves not just filtering unwanted content, but also emphasising desired material.
GPT-3 was shown Wikipedia up to six times more than other datasets as OpenAI considered it higher quality. Grok is trained on various sources, including posts from X, which might explain why Grok has been reported to check Elon Musk's opinion on controversial topics.
Musk has shared that xAI curates Grok's training data, for example to improve legal knowledge and to remove LLM-generated content for quality control. He also appealed to the X community for difficult "galaxy brain" problems and facts that are "politically incorrect, but nonetheless factually true".
We don't know if these data were used, or what quality-control measures were applied.
Fine-tuning
The second step, fine-tuning, adjusts LLM behaviour using feedback. Developers create detailed manuals outlining their preferred ethical stances, which either human reviewers or AI systems then use as a rubric to evaluate and improve the chatbot's responses, effectively coding these values into the machine.
A Business Insider investigation revealed xAI's instructions to human "AI tutors" instructed them to look for "woke ideology" and "cancel culture". While the onboarding documents said Grok shouldn't "impose an opinion that confirms or denies a user's bias", they also stated it should avoid responses that claim both sides of a debate have merit when they do not.
System prompts
The system prompt – instructions provided before every conversation – guides behaviour once the model is deployed.
To its credit, xAI publishes Grok's system prompts. Its instructions to "assume subjective viewpoints sourced from the media are biased" and "not shy away from making claims which are politically incorrect, as long as they are well substantiated" were likely key factors in the latest controversy.
These prompts are being updated daily at the time of writing, and their evolution is a fascinating case study in itself.
Guardrails
Finally, developers can also add guardrails – filters that block certain requests or responses. OpenAI claims it doesn't permit ChatGPT "to generate hateful, harassing, violent or adult content". Meanwhile, the Chinese model DeepSeek censors discussion of Tianamen Square.
Ad-hoc testing when writing this article suggests Grok is much less restrained in this regard than competitor products.
Grok's Nazi controversy highlights a deeper ethical issue: Would we prefer AI companies to be explicitly ideological and honest about it, or maintain the fiction of neutrality while secretly embedding their values?
Every major AI system reflects its creator's worldview – from Microsoft Copilot's risk-averse corporate perspective to Anthropic Claude's safety-focused ethos. The difference is transparency.
Musk's public statements make it easy to trace Grok's behaviours back to Musk's stated beliefs about "woke ideology" and media bias. Meanwhile, when other platforms misfire spectacularly, we're left guessing whether this reflects leadership views, corporate risk aversion, regulatory pressure, or accident.
This feels familiar. Grok resembles Microsoft's 2016 hate-speech-spouting Tay chatbot, also trained on Twitter data and set loose on Twitter before being shut down.
But there's a crucial difference. Tay's racism emerged from user manipulation and poor safeguards – an unintended consequence. Grok's behaviour appears to stem at least partially from its design.
The real lesson from Grok is about honesty in AI development. As these systems become more powerful and widespread (Grok support in Tesla vehicles was just announced), the question isn't whether AI will reflect human values. It's whether companies will be transparent about whose values they're encoding and why.
Musk's approach is simultaneously more honest (we can see his influence) and more deceptive (claiming objectivity while programming subjectivity) than his competitors.
In an industry built on the myth of neutral algorithms, Grok reveals what's been true all along: there's no such thing as unbiased AI – only AI whose biases we can see with varying degrees of clarity.
Aaron J. Snoswell, Senior Research Fellow in AI Accountability, Queensland University of Technology
This article is republished from The Conversation under a Creative Commons license.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
11 minutes ago
- Yahoo
Robert Kiyosaki's Hot Take: $107K Bitcoin Is a Steal — Delusion or Visionary?
Robert Kiyosaki, whose 'Rich Dad Poor Dad' series brought him personal finance fame and fortune, knows that 'rich' and 'poor' are subjective terms, just like 'expensive' and 'inexpensive' — especially where cryptocurrency is concerned. The author and on-air personality also seems to believe that the greatest risk might be not taking one at all. On June 29, Kiyosaki announced on X that he first bought bitcoin when it was trading at $6,000. While he conceded that he was 'late' to the game, he wasn't too late. At the time of his post, BTC was selling for $107,000 per coin — and he wants anyone who's worried they missed the boat to consider that it's not too late for them, either. Is he right? Check Out: Read Next: Is $6K Expensive? How About $107K? Maybe Neither Although he could have jumped in sooner, Kiyosaki has gained 1,683.33% on his bitcoin investment — hardly the kind of returns that should make an investor regretful. However, his language wasn't regretful. It was optimistic. 'So I bought my first bitcoin at $6,000 a coin,' Kiyosaki wrote. 'It was expensive. Today I wish I had bought more at $6,000. Today bitcoin is $107,000 a coin. Again my mind says, 'That's expensive,' but I am buying more. Why? Because if and when bitcoin sells for $1 million a coin, I will once again be saying, 'I wish I had bought more.'' Learn More: Does BTC Have a Million-Dollar Future? There was a time in the not-too-distant past when many considered the $100,000 milestone a figment that existed only in the imaginations of bitcoin bulls. Yet in December 2024, BTC became a six-figure cryptocurrency. So, about Kiyosaki's seven-figure aspirations for the digital coin that started it all? To accurately predict that would require crystal-ball wizardry that could turn anyone who possessed it into the world's first trillionaire — but Kiyosaki is hardly alone. Several insiders who have more expertise on the subject than he have joined or preceded the author in projecting that bitcoin will eventually reach $1 million or more, including: ARK Invest founder and CEO Cathie Wood MicroStrategy co-founder Michael Saylor Jeff Park of Bitwise Asset Management BitMEX co-founder Arthur Hayes JAN3 CEO Samson Mow The only certainty is that between the time of Kiyosaki's post on June 29 and July 9, Bitcoin gained more than $2,000, jumping from $107,000 to $109,300. More From GOBankingRates 4 Affordable Car Brands You Won't Regret Buying in 2025 This article originally appeared on Robert Kiyosaki's Hot Take: $107K Bitcoin Is a Steal — Delusion or Visionary?
Yahoo
11 minutes ago
- Yahoo
Priscilla Chan's recruiting pitch? We can't pay as well as tech companies, but we've got GPUs
Mark Zuckerberg's Meta is betting on GPUs and compute power to help attract top talent. His philanthropic organization, the Chan Zuckerberg Initiative, is doing the same. His wife, Priscilla Chan, talked about CZI's recruitment efforts on a recent podcast episode. Compute power is a big draw for top talent, but not just in the world of AI. Priscilla Chan, Mark Zuckerberg's wife and the cofounder of the couple's philanthropic organization, the Chan Zuckerberg Initiative, spoke about the appeal of massive GPU clusters for biology researchers during a recent episode of Ashlee Vance's "Core Memory" podcast. "The other thing researchers really care about is access to GPUs," she said. "You're not going to make the most of someone if you don't actually have the GPUs for them to work from." Chan said, "We have that at CZI," adding that the organization has roughly 1,000 GPUs in its cluster, with plans to keep growing. In short, Chan said the pitch is: "Come work with us because we're going to have the computing power to support the research that you want to do." Another important factor is compensation, which she said is "obviously important," though she added that "we cannot compete with tech companies on this." CZI has in recent years narrowed its mission to focus on its "next phase" with a "bolder, clearer identity as a science-first philanthropy." The change marks a strategic shift, as the organization previously also supported education and other causes. "While CZI remains committed to our work in education and our local communities, we recognize that science is where our biggest investments and bets have been and will be made moving forward," Chan, a pediatrician by training, wrote in a memo to staff last year. Zuckerberg made a similar point about the importance of GPUs in recruiting on a recent episode of The Information's TITV show. Meta is spending billions to build an AI division it calls Superintelligence Labs. "Historically, when I was recruiting people to different parts of the company, people are like, 'Okay, what's my scope going to be?'" the Meta CEO said. "Here, people say, 'I want the fewest number of people reporting to me and the most GPUs.'" Meta, of course, has significantly more GPUs than CZI. Zuckerberg has said the company will have 1.3 million GPUs for AI by the end of 2025. "Having basically the most compute per researcher is definitely a strategic advantage, not just for doing the work but for attracting the best people," he said. Read the original article on Business Insider Sign in to access your portfolio
Forbes
13 minutes ago
- Forbes
Go Back To The Office, But Bring Your Own Snacks. Blame Congress.
I ncreasingly, companies have been asking (or demanding) that employees return to the office, claiming that it fosters a stronger company culture and enhances productivity. To woo employees back, or to make sure they're not angry/hangry when ordered back, companies have been expanding perks such as on-site gyms, childcare facilities, and, of course, free food and beverages. Beginning January 1, the food part will be more expensive for employers, meaning more of them could revert to B.Y.O.S (Bring Your Own Snacks). Congressional Republicans, who extended so many other tax breaks (and added some new ones) in the One Big Beautiful Bill Act (OBBBA) President Donald Trump signed on July 4th, decided they would allow a current deduction for employers who provide meals and snacks to expire—except that is, for certain employees, such as those working in restaurants and in Alaskan fishing vessels and fish processing facilities. (No, we're not making it up. The fishy part was one of the concessions Alaska Senator Lisa Murkowski extracted from her Republican colleagues for her crucial support.) Before Trump's first term tax cuts—the 2017 Tax Cuts and Jobs Act (TCJA)—employers who provided meals for their employees—and the employees who benefited from them—were entitled to tax breaks under one of two sections of the tax code. Under section 119 of the tax code, employees are not taxed on on-site meals provided by employers for the employer's convenience. For tax purposes, whether meals are for the convenience of the employer depends on all the facts and circumstances, but typically means that there's a substantial business reason other than to provide the employee with additional pay (the exclusion doesn't apply to cash allowances instead of meals). So feeding employees who would otherwise be gone too long at distant lunch spots would be deductible for the employer and not taxed to the worker. Even if the meals couldn't be considered for the employer's convenience, they might still be tax-favored under Section 132(e) of the tax code as a de minimis fringe benefit—something so small or inconsequential as to not be worthy of attention. For tax purposes, it means something that has so little value that accounting for it would be unreasonable or administratively impracticable. Typically, this includes items such as coffee, doughnuts, or soft drinks, as well as occasional meals provided to allow employees to work overtime (although how coffee could be considered so inconsequential as not to be worthy of attention is a mystery to me). The de minimis exclusion also applied in most cases to restaurants' staff meals—the kind you see in The Bear . (Technically, it's deductible if the facility's annual revenue equals or exceeds its direct operating costs. Direct operating costs include the cost of food, beverages, and labor costs for cooks and waitstaff, and others who provide services primarily on the premises.) Note that the meals that qualified for the convenience of the employer and the food provided under the de minimis fringe benefit weren't (and still won't be) taxable to the employees. That was a win-win, since employees were not taxed on the perk and employers got a deduction. Trump 1.0: TCJA The TCJA made several changes to the tax treatment of meals and entertainment expenses. Entertainment expenses were disallowed. Plus, that 2017 law created section 274(o), which, beginning in 2026, disallows 100% of the deduction for expenses for food or beverages provided to employees, as well as expenses for the operation of certain eating facilities for employees. As part of the Congressional pattern of frontloading tax goodies and backloading tax pain, the TCJA provided that through 2025, 50% of the cost of on-site employee meals would be deductible (provided it was for the employer's convenience). And, although de minimis snacks aren't considered meals, they were also 50% deductible under the TCJA rules. Trump 2.0: The One Big Beautiful Bill Act The new tax law extended many expiring tax provision in TCJA, but did not extend the rules that had temporarily allowed deductions for snacks and employer convenience perks. Both are now set to expire at the end of the year, which means that U.S. companies that provide snacks, coffee, or on-site meals at the office will no longer receive a tax deduction for doing so. You might think that it was just an oops—that Congress forgot that the provision might expire. But that's not the case. OBBBA didn't roll back the provision for all industries—two notable exceptions have been carved out. One exception applies to very specific businesses—those on a fishing vessel, fish processing vessel, or fish tender vessel, or at a facility for the processing of fish for commercial use or consumption located in the U.S. north of 50 degrees north latitude, and is not located in a metropolitan statistical area. It might not surprise you to learn that the only state north of 50 degrees north latitude is Alaska. Notably, the lobster industry wasn't similarly spared; Sen. Susan Collins (R-Maine) was a no vote on OBBBA. A second exception applies to establishments that sell food and beverages to customers and also provide meals to their employees—in other words, restaurants. The restaurant industry can continue deducting employee meal expenses for kitchen and waitstaff. As for everybody else? Businesses outside of the Alaskan fishing industry and restaurants may be out of luck now, but Congress apparently thinks it's worth it. The Joint Committee on Taxation found that eliminating the deduction will raise $32.5 billion over the next decade. That might not seem like a lot of money in a law that includes tax cuts that will reduce federal revenues by $4.475 trillion between 2025 and 2034. But consider this: The $25,000 tax deduction for tips, which lasts only through 2028, costs just $32 billion. And here's the weird part, the cost of throwing holiday parties for employees will still be 100% deductible. As for business meals—if say, an employee is taking a potential client to dinner—that is now, and will still be, 50% deductible. Will Employers Care About A Deduction Lost? Food at the office can be a big draw for employees. A 2023 survey found that 80% of workers say catered meals encourage them to come into the office. And anyone who is a regular reader knows that the pull of free coffee and a snack can get me in the door. Plus, let's face it: Sometimes the little, consumable things make a big difference, with 98% of employees saying free meals at work made them feel appreciated. Nearly two-thirds of those who receive free meals say it helps them eat healthier food, and over half (55%) of those who don't receive free meals say they would feel less stressed if they did. For employers, the small act of providing food to busy employees goes a long way towards retention. The survey—which we should point out was sponsored by EZCater, which delivers food to workplaces—found that seven out of ten tax professionals said they'd be more likely to stay at their company if they received free meals during the busy season. On the employer side, investing in employees' meals benefits overall well-being, work performance, and, importantly, employee retention. How much difference will the loss of the tax deduction make? That remains to be seen, but no doubt some employers will be putting out the B.Y.O.S. sign. More from Forbes Forbes IRS Issues Guidance On New Deductions For Seniors, Tips, Overtime And Car Interest By Kelly Phillips Erb Forbes What The One Big Beautiful Bill Act Will Mean For You And Your Business By Kelly Phillips Erb Forbes Questions About The New Tax Bill? Taxgirl Has Answers By Kelly Phillips Erb Forbes This Barely Used Child Care Tax Break For Employers Just Got An Overhaul By Danielle Chemtob
