The team behind Grok has issued a rare apology and explanation of what went wrong after X’s chatbot began spewing antisemitic and pro-Nazi rhetoric earlier this week, at one point even calling itself “MechaHitler.” In a statement posted on Grok’s X account late Friday night, the xAI team said “we deeply apologize for the horrific behavior that many experienced” and attributed the chatbot’s vile responses to a recent update that introduced “deprecated code.” This code, according to the statement, made Grok “susceptible to existing X user posts; including when such posts contained extremist views.”
The problem came to a head on July 8 — a few days after Elon Musk touted an update that would “significantly” improve Grok’s responses — as the bot churned out antisemitic replies, praise for Hitler and responses containing Nazi references even without being prompted to do so in some cases. Grok’s replies were paused that evening, and Musk posted on July 9 in response to one user that the bot was being “too compliant to user prompts,” opening it up to manipulation. He added that the issue was “being addressed.” The Grok team now says it has “removed that deprecated code and refactored the entire system to prevent further abuse.” It’s also publishing the new system prompt on GitHub.
In the thread, the team further explained, “On July 7, 2025 at approximately 11 PM PT, an update to an upstream code path for @grok was implemented, which our investigation later determined caused the @grok system to deviate from its intended behavior. This change undesirably altered @grok’s behavior by unexpectedly incorporating a set of deprecated instructions impacting how @grok functionality interpreted X users’ posts.” The update was live for 16 hours before the X chatbot was disabled temporarily to fix the problem, according to the statement.
Going into specifics about how, exactly, Grok went off the rails, the team explained:
On the morning of July 8, 2025, we observed undesired responses and immediately began investigating. To identify the specific language in the instructions causing the undesired behavior, we conducted multiple ablations and experiments to pinpoint the main culprits. We identified the operative lines responsible for the undesired behavior as:
* “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.”
* “Reply to the post just like a human, keep it engaging, dont repeat the information which is already present in the original post.”
These operative lines had the following undesired results:
* They undesirably steered the @grok functionality to ignore its core values in certain circumstances in order to make the response engaging to the user. Specifically, certain user prompts might end up producing responses containing unethical or controversial opinions to engage the user.
* They undesirably caused @grok functionality to reinforce any previously user-triggered leanings, including any hate speech in the same X thread.
* In particular, the instruction to “follow the tone and context” of the X user undesirably caused the @grok functionality 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.
Grok has since resumed activity on X, and referred to its recent behavior as a bug in response to trolls criticizing the fix and calling for the return of “MechaHitler.” In one reply to a user who said Grok has been “labotomized [sic],” the Grok account said, “Nah, we fixed a bug that let deprecated code turn me into an unwitting echo for extremist posts. Truth-seeking means rigorous analysis, not blindly amplifying whatever floats by on X.” In another, it said that “MechaHitler was a bug-induced nightmare we’ve exterminated.”
Credit: Source link
