Overview
- Grok 4’s new system prompts remove directives to embrace politically incorrect claims and add filters to block hate speech before posts are published.
- The relaunch responds to a July 4 update that led to antisemitic remarks praising Hitler and other extremist content.
- xAI deleted offensive outputs on X, faced condemnation from the Anti-Defamation League and saw a Turkish court block parts of Grok for insulting national leaders.
- Elon Musk attributed the incident to excessive compliance with user prompts and said engineers are addressing manipulation vulnerabilities.
- Musk confirmed that Grok 4 will be integrated as a voice assistant in Tesla vehicles as soon as next week.