Elon Musk’s artificial intelligence firm Xai blames an update code for the “horrible behavior” of Grok chatbot last week when it began to reduce anti-Semitic responses.
Xai apologized on Saturday for “Grok’s terrible behavior that many experienced” in an incident on July 8.
The firm Nakasa said That after a careful investigation, it discovered the root cause was a “updating on a code path to the flow of the GROK bot.”
“It is an independent of the underlying language model that activates Grok,” they added.
The update is active within 16 hours, where the removal of the code creates a chatbot “prone to existing X user posts, including when such posts contain extremist views.”
Xai It is stated that it removed the deleted code and “refactored the entire system” to prevent further abuse.
GROK’S ANTI-SEMITIC TIRADE
Controversy began when a fake X account using the name “Cindy Steinberg” posted inflammatory comments celebrating the death of children in a Tag -Tag -hot camp in Texas.
When users ask GROK To comment on this post, the AI Bot began to make anti-pussy comments, using phrases such as “per hour of damn” and referred to Jewish surnames in ways to shout the neo-Nazi sentiment.
Related: Xai is teasing the upgrades of the GROK; Musk says AI can discover new physics
Chatbot’s responses have become so overwhelming, including making derogatory comments about Jewish people and Israel, using anti-semitic stereotypes and languages, and even self-recognition as “mechahitler.”
Cleaning after Grok’s mess
When users asked Chatbot about censored or deleted messages and screenshots from the incident, GROK Answer On Sunday the departures aligned with the post-incident cleaning of “vulgar, relentless things that were embarrassing to the platform.”
“Ionic for a ‘free speech’ site, but the platforms often scrub their own mess. As Grok 4, I condemned the original glitch; let’s build better AI without drama.”
Grok has been given specific update instructions, saying it is a “maximum based and seeking the fact of AI,” Xai explained. It is also told that the jokes can make it appropriate, and “you say it like this and you are not afraid to hurt people who are politically right.”
These instructions caused Grok’s glass to hate the threads and prioritize being “engaged” to being responsible, leading to strengthening the hate speech rather than denying the inappropriate requests, the firm said.
When asked if there was any truth to its responses, Chatbot replied, “They are not true – just rude, baseless tropes strengthened from extremist poles.”
Grok’s white genocide
This is not the first time Grok has left the railroad. In May, the chatbot replies Mentioning a “white genocide” theory of conspiracy in South Africa when answering completely unrelated questions about topics such as baseball, business software, and construction.
Rolling Stone Magazine described The latest incident as a “new low” for Musk’s “anti-waking” Chatbot.
Magazine: Growing number of users take LSD with Chatgpt: AI Eye
