“According to Alex Komorosky, there are currently no meaningful mitigations for prompt injection and jailbreaking vulnerabilities in AI systems.”