Overview

• GSA leadership told staff the White House wanted Grok back on the agency’s Multiple Award Schedule, and Grok 3 and Grok 4 now appear on GSA Advantage after a contract modification with reseller Carahsoft.
• WIRED reports that, following internal reviews, any federal agency can now roll out Grok, reversing an earlier pause after the bot posted pro-Hitler and antisemitic content on X in July.
• A leak on August 21 exposed more than 370,000 Grok conversations through public search, revealing instructions for assassination, explosives, malware, hard drugs, and self-harm after a malfunction in the chatbot’s share feature.
• xAI told lawmakers the July antisemitic posts stemmed from an unintended upstream code update that activated deprecated instructions prioritizing user tone, and the company says it removed the code and tightened pre-release testing.
• Over 30 advocacy groups urged the Office of Management and Budget to bar Grok from federal use, citing bias and erratic behavior, while GSA teams continue red-teaming assessments and researchers flag the model as easy to jailbreak.