After the release of Claude Mythos Preview, a notable spike in severe vulnerabilities was observed. The incidents highlight the need for rigorous security vetting during product rollouts. These findings underscore potential risks in AI model deployments.