Anthropic's Claude back globally after two-week US government ban over jailbreak

The US government has lifted a two-week ban on Anthropic shipping its AI model globally, imposed after Amazon researchers discovered a jailbreak exploit. Anthropic noted that even smaller models like Claude Haiku 4.5 are vulnerable to the same technique. A new safety classifier now blocks the exploit in over 99% of cases, though it also incorrectly flags some harmless requests.