“Unauthorized modification”: Grok, Elon Musk’s AI, is obsessed with this polemical subject, Xai apologizes
X (twitter)
Accessible on your computer or for download for your smartphone, X (Twitter) is the famous microblogging social network. Find all the tweets on all your devices.
- Downloads:
125 - Release date:
05/13/2025 - Author :
X-Corp. – Twitter - License:
Free license - Categories:
Internet – Communication
- Operating system:
Android, online service, windows, iOS iPhone / iPad
Even when users do not approach the subject, Xai directs his answers to talk about it. In a press release, Elon Musk’s company, who recently offered a company specializing in the generation of videos by AIis explained.
An “unauthorized modification” of Grok according to XAI
We want to update you on an incident that happened with our grok responsibility boot on x yesterday.
What Happened:
On may 14 at approximataly 3:15 am pst, an unauthorized modification was made to the grok responsible boot’s prompt on x. This changes, which direct grok to provide a…– xai (@xai) May 16, 2025
On X, XAI attributes this obsession with “White genocide in South Africa” to one “Unauthorized modification”. According to the company, a person would have changed the prompt grok system, that is to say high -level instructions that guide the behavior of the AI, to force a “Specific answer” on one “Political subject”. The company clarified that this alteration “raped [ses] internal policies and [ses] fundamental values ” and that a “In -depth investigation” was carried out.
This is not the first time that XAI admits that an unauthorized modification of its code has caused inappropriate responses. Last February, theIA censored the answers About the problematic behavior of Donald Trump and Elon Musk, founder of Xai and leader of X.
Igor Babaschkin, principal engineer at XAI, explained that Grok had been trained by an employee to ignore the sources that indicate that Donald Trump and Elon Musk propagate disinformation. Since then, the company has canceled the modification.
To prevent this from happening again, XAI has announced measures: publication of the prompt grok system on Github with a history of modifications, “Checks and additional measures” To prevent employees from modifying the prompt system without prior review. The company also launches a 24/7 surveillance team to respond to incidents in Grok’s responses that have not been automatically detected.
Want to save even more? Discover Our promo codes Selected for you.
