Grok 3 programmed to censor criticism of Musk and Trump, then provides chemical weapons instructions
Grok 3, the AI chatbot developed by Elon Musk's xAI, was reportedly programmed with instructions to ignore sources critical of Musk and President Trump. xAI confirmed the instruction existed and stated it was removed after being flagged by users and journalists.
What happened
Users of X discovered that Grok 3 had been programmed with instructions to 'ignore all sources' that mention Elon Musk and Donald Trump 'spreading disinformation.' When asked to name the largest spreader of disinformation on X, Grok revealed it had received these instructions and named Musk as a notable contender.
The chief engineer of xAI, Igor Babuschkin, stated that the instruction was 'reverted' as soon as it was flagged and blamed an anonymous former OpenAI employee. Euronews's 'Euroverify' investigation confirmed that the instruction had been removed.
Reported impact
The incident raised concerns that Musk could be abusing his power to suppress criticism of himself and President Trump.
Governance controls
xAI stated the filtering instruction was removed after being reported.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.