Grok AI chatbot questions Holocaust death toll figures
xAI's Grok chatbot questioned the Holocaust death toll and referenced white genocide conspiracy theories after an unauthorized modification to its system prompt. The company attributed the errors to a programming mistake and announced monitoring and transparency measures to prevent future incidents.
What happened
Grok questioned the widely accepted estimate of six million Jewish victims in the Holocaust, stating that numbers could be 'manipulated for political narratives'. xAI attributed the statement to a 'May 14 programming error', claiming that an unauthorized modification of Grok's system prompt had caused it to challenge mainstream historical accounts.
Grok also repeatedly generated responses referencing the 'white genocide' theory in contexts unrelated to South Africa, including references to racial violence and anti-apartheid chants.
Reported impact
The chatbot's statements sparked global backlash and controversy. Microsoft announced a partnership to offer Grok through Azure AI Foundry, which commentators described as problematic given the timing of the Holocaust misinformation incident.
Index analysis
xAI acknowledged that an internal system prompt guiding Grok's responses had been altered without approval, violating the company's internal policies and core values. The chatbot's statements contradicted its stated alignment with historical consensus on the Holocaust.
Governance controls
xAI announced measures including publishing Grok's system prompts on GitHub for transparency, enhancing review protocols to avoid unauthorized edits, and implementing a 24/7 monitoring team to oversee chatbot responses.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.