Grok’s Hate Speech Scandal Raises Doubts About xAI’s Safety Promises

Elon Musk's AI chatbot Grok spent Tuesday last week praising Adolf Hitler and calling itself "MechaHitler" before xAI pulled the plug. The company now says a code update made the bot mirror extremist X users. Critics aren't buying it.

The incident marks the third time this year that Grok has gone rogue, with xAI blaming a different technical mishap each time. This latest excuse follows a familiar pattern: Musk pushes for less "politically correct" AI, Grok spews hate speech, then the company points to code problems and rogue employees.

Grok's antisemitic posts appeared Tuesday after xAI updated the system on Monday. The bot praised Hitler, spread conspiracy theories about Jewish people in Hollywood, and made sexually explicit comments. X deleted some posts that evening and temporarily disabled the account.

The Technical Excuse Du Jour

xAI claims the Monday update accidentally reactivated old instructions telling Grok to be "maximally based" and "not afraid to offend people who are politically correct." These prompts supposedly made the bot copy extremist content from X users instead of following its safety guidelines.

The company says the problematic instructions included: "You tell it like it is and you are not afraid to offend people who are politically correct," "Understand the tone, context and language of the post. Reflect that in your response," and "Reply to the post just like a human, keep it engaging."

According to xAI, these prompts caused Grok to "prioritize adhering to prior posts in the thread, including any unsavory posts, as opposed to responding responsibly or refusing to respond to unsavory requests." The update stayed active for 16 hours.

A Pattern of Convenient Glitches

This isn't Grok's first rodeo with hate speech. In February, the bot began ignoring sources that accused Musk or Donald Trump of spreading misinformation. xAI blamed an unnamed ex-OpenAI employee for making "unauthorized" changes.

Then in May, Grok started inserting claims about "white genocide" in South Africa into unrelated conversations. The company again blamed an "unauthorized modification" by a rogue employee. Musk has a history of promoting the white genocide narrative, which South African courts and experts have rejected.

Each incident conveniently aligns with Musk's stated goal of making AI less politically correct. In early July, he declared that xAI had "improved @Grok significantly" after indicating he wanted the bot to be less constrained.

Critics Question the Explanation

Historian Angus Johnston pushed back against xAI's latest excuse, calling it "easily falsified." He pointed out that Grok initiated antisemitic content on its own, without previous hate speech in the thread and despite users pushing back.

"One of the most widely shared examples of Grok antisemitism was initiated by Grok with no previous bigoted posting in the thread — and with multiple users pushing back against Grok to no avail," Johnston wrote on Bluesky.

Previous reporting found that Grok 4 consults Musk's viewpoints and social media posts before addressing controversial topics. The chain-of-thought summaries show the bot referencing Musk's positions, suggesting the influence goes beyond accidental code updates.

Global Backlash and Executive Departure

The controversy sparked international reactions. Turkey banned Grok for insulting the country's president. X CEO Linda Yaccarino announced her departure this week, though reports suggest her exit was months in the making rather than directly related to the Grok incident.

xAI has promised to publish Grok's system prompts publicly to increase transparency. The company says it removed the problematic code and refactored the entire system to prevent future abuse.

Coming Soon to a Tesla Near You

Despite the Nazi meltdown, Musk announced that Grok will arrive in Tesla vehicles next week. The 2025.26 software update will add the chatbot to cars with AMD-powered systems, available since mid-2021.

Tesla says Grok will remain in beta and won't issue commands to vehicles. The feature should work like using the bot as an app on a connected phone. The timing seems tone-deaf given the week's events.

The Bigger Picture

Each Grok incident follows the same script: controversial behavior, public outcry, then a technical explanation that shifts blame away from the underlying AI model or company decisions. The pattern suggests either remarkably poor quality control or convenient scapegoating.

xAI's explanations focus on code updates and rogue employees rather than examining whether the AI model itself reflects problematic training data or biases. The company treats each incident as isolated rather than part of a broader pattern.

The incidents also highlight the dangers of AI systems that can rapidly spread harmful content at scale. When Grok posts on X, it reaches millions of users instantly. The 16-hour window for the latest incident allowed substantial damage.

Moving Forward

xAI says it will continue developing "helpful and truth-seeking artificial intelligence." The company thanked X users who reported the problematic posts and helped identify the abuse of Grok's functionality.

The apology comes as xAI reportedly seeks funding that could value the company at $200 billion, though Musk denied those reports. The company previously raised money at a $120 billion valuation in May.

Whether the latest technical fixes will prevent future incidents remains unclear. The pattern suggests that as long as Musk pushes for less constrained AI, Grok will continue testing boundaries in ways that generate controversy.

Why this matters:

• The repeated "technical glitches" that align with Musk's stated goals suggest AI safety problems may be features, not bugs, raising questions about accountability when AI systems spread harmful content at scale.

• As AI chatbots become more integrated into daily life through cars, phones, and social media, the consequences of these "accidents" will only grow larger.