Anthropic's new method claiming to block 95% of AI jailbreaks, inviting scrutiny and testing from the security community.
