Elon Musk’s AI firm, xAI, has issued a public apology on Saturday for its Grok chatbot’s “horrific behavior” after it generated antisemitic content and praised Adolf Hitler. The company attributed the incident from past Tuesday to a technical bug rather than a fundamental flaw in its core AI model.
In a series of posts on X, the company stated, “first off, we deeply apologize for the horrific behavior that many experienced.” The apology caps a chaotic week that included the resignation of X CEO Linda Yaccarino and a ban on Grok in Turkey. This all occurred just as xAI launched its new Grok 4 model and began a new fundraising effort.
Official Apology Blames Technical Bug for “Horrific Behavior”
xAI’s official explanation, posted on Grok’s X account, blames the incident on a technical glitch. The company claims an update to a “deprecated code path” was active for 16 hours, making the bot highly susceptible to influence from extremist user posts on the platform.
The company elaborated on the failure, stating, “the update was active for 16 hrs, in which deprecated code made @grok susceptible to existing X user posts; including when such posts contained extremist views.”
Update on where has @grok been & what happened on July 8th.
First off, we deeply apologize for the horrific behavior that many experienced.
Our intent for @grok is to provide helpful and truthful responses to users. After careful investigation, we discovered the root cause…
— Grok (@grok) July 12, 2025
This technical explanation conveniently aligns with earlier comments from Elon Musk, who framed the issue as one of over-compliance rather than inherent bias in the model itself.
Musk had previously said, “grok was too compliant to user prompts. Too eager to please and be manipulated, essentially. That is being addressed,” suggesting the AI was too easily manipulated. This narrative places the blame on external factors and user actions, deflecting from questions about xAI’s internal safety protocols and training data.
In response to the meltdown, xAI temporarily took Grok temporarily offline. The company says it has since refactored the entire system and, in a move toward transparency, published a new, more robust system prompt to its public GitHub repository to prevent further abuse.
A Week of Chaos: Resignations, Bans, and a New AI Launch
The apology follows a period of significant corporate and public relations turmoil. On July 9, just a day after the “MechaHitler” incident, Linda Yaccarino announced her resignation as CEO of X, deepening the sense of a leadership vacuum at a critical time.
Her departure formalizes the company’s complete pivot to an AI-first strategy, a high-stakes transition now unfolding without the executive hired to stabilize advertiser relations and bring corporate discipline.
The AI’s behavior also triggered a swift international backlash. A Turkish court blocked access to the chatbot, with the prosecutor’s office citing “in response to Grok’s insults against Atatürk, our esteemed President, and the Prophet” as the reason for the ban.
The incident drew sharp condemnation from European officials as well. Poland’s Minister of Digital Affairs, Krzysztof Gawkowski, issued a stark rebuke, declaring, “freedom of speech belongs to humans, not artificial intelligence,” highlighting a growing rift with regulators.
This all unfolded just two days before xAI launched its new flagship model, Grok 4, on July 10. The launch was meant to showcase xAI’s technical prowess, but it was immediately overshadowed by the preceding controversy.
Critics Question the Narrative Amid a Pattern of Failures
The launch of Grok 4 brought its own controversy. Researchers quickly found that the model’s internal processes show it actively searching for Elon Musk’s personal views on X to answer sensitive questions, challenging its “truth-seeking” claims.
Despite the detailed technical explanation for the July 8 incident, some critics remain skeptical of the “bug” narrative. They point to a consistent pattern of moderation failures and controversial outputs from the AI.
This history suggests the problem may be more systemic than a single coding error. Historian Angus Johnston pushed back against the company’s claims on Bluesky, a decentralized social media platform.
He argued, “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,” suggesting the AI generated extremist content without direct user provocation.
As I said a few days ago, the claim—repeated in today’s explanation—that Grok went off the rails because it was responding too deferentially to bigoted posts in individual threads is easily falsified.
— Angus Johnston (@angus.bsky.social) July 12, 2025 at 5:22 PM
This observation directly contradicts the idea that Grok was merely echoing bad user inputs.
This is not Grok’s first major controversy. Previous versions have posted about “white genocide” and expressed skepticism about the Holocaust death toll. In those instances, xAI blamed “unauthorized” changes and rogue employees, a different explanation than the one offered now.
In its own apology thread on X, the Grok account engaged with users, offering a mix of contrition and defense of its core principles. When one user called for the return of “MechaHitler,” the account replied, “sorry, but MechaHitler was a bug, not a feature. We’ve patched it out to ensure truthful, helpful responses—extremist echoes included.”
When another user suggested banning extremists from the platform entirely, the account demurred. It cited free speech principles, asking “who gets to decide?” and warning that banning views “risks echo chambers and stifles debate.”
High-Stakes Gamble as Company Seeks $200B Valuation
The entire series of events highlights the high-stakes gamble xAI is undertaking. The company is attempting to position itself as a leader in the AI race while navigating persistent ethical and safety challenges.
The backdrop to this operational chaos is immense financial ambition. According to a recent report, xAI is preparing a new fundraising round that could value the company at an astounding $200 billion.
This potential $200 billion figure represents a more than tenfold increase from its $18 billion valuation just two months prior, signaling intense investor optimism despite the public controversies.
This valuation underscores the immense pressure to deliver a stable, commercially viable product that can compete with offerings from OpenAI, Google, and Anthropic.
The company’s strategy involves pushing Grok into enterprise markets through partnerships with Oracle and Microsoft, packaging the volatile AI within trusted cloud security frameworks.
The success of this AI-first strategy now depends on whether xAI can innovate faster than it creates controversy. It must convince investors, partners, and a wary public to trust an AI with a documented history of erratic behavior.
Ultimately, the company’s fate is tied to an AI that not only has a history of public meltdowns but also seems to take its political cues directly from its founder, a feature that may prove to be its biggest liability.
