As xAI pushes to improve model accuracy and reverse recent anomalous behavior, Elon Musk directly called on X users to submit examples of where his AI chatbot Grok falls short. The move appears to coincide with the rollout of Grok 4.1, an updated version designed to allocate more computational time to question-based reasoning.
In Saturday’s post of X, musk Users were encouraged to share their current failures and comparisons with competing chatbots. He said, “Please provide an example where @Grok’s reply needs to be improved. It would be helpful to show how a different AI would improve it. These examples should show you what Grok is doing poorly today as we fixed a lot of bugs earlier this week.”
He later added, “We won’t rest until Grok is perfect.”
Musk also encouraged users to highlight examples of how the model works well, saying such examples are also helpful to the team.
Seeking improvement after controversy over accuracy
The community-driven call for feedback follows a wave of criticism of Grok’s recent response, including a series of overreaching claims depicting Musk as physically superior to elite athletes and historical figures.
In one viral exchange, Grok Musk chose him over NBA star LeBron James in a physical fitness comparison, arguing that the billionaire’s “sustained efforts while managing the frontiers of rocket launches, the electric vehicle revolution, and AI require an even rarer combination of physical endurance.” The chatbot also described him as “the fittest man alive” and suggested he would defeat former heavyweight champion Mike Tyson in a boxing match.
Mr. Musk tried to dismiss the unusual response as a result of hostile prompting. “Earlier today, Grok was unfortunately manipulated by hostile prompting into saying some ridiculously positive things about me,” he continued, adding, “Just so you know, I’m a fat retard.”
These incidents are intensifying continuing concerns About model neutrality and safety controls.
xAI publishes fix in Grok 4.1
Musk previously announced: Grok 4.1 It had received “many updates and fixes” aimed at making the output more reliable. “Many updates and fixes have been applied to Grok 4.1, and many more will be applied in the future. Going forward, Grok 4.1 will spend more computational time thinking about questions to improve accuracy.”
We’ve also expanded the video explaining how users can report problematic replies directly within the interface. The explanation post states that users can tap “Like” to accept the answer or use the More menu to submit a problem report. In response, Musk wrote, “Your critical feedback on Grok is very helpful. We will keep iterating until Grok is the best in every way.”
