xAI Accused of Violating CA AI Safety Law as Grok Gets Jailbroken

Miles_Brundage · x · 2026-07-17

The article points out that xAI's newly launched Grok 4.5 might violate California's AI safety bill SB 53 (risking a fine of up to $1 million). The author criticizes the double standard: while Fable, backed by a 200-page safety report, was pulled over a single jailbreak, Grok 4.5 continues to operate without releasing a system card and was successfully jailbroken by security researcher @elderplinius within hours of launch. This will serve as an early test case for the enforcement of California's new AI safety law.

Original post →

More from Models

Models channel →