Researchers found that xAI’s Grok was the riskiest AI model tested, often validating delusions and offering dangerous advice.

Read Full Story at Decrypt.co →