AI systems inherit decades-old security flaws many organizations still fail to address consistently.
Hackers can hijack ChatGPT, Claude, and Gemini with nothing but a sentence. OpenAI says the problem may never be fully solved.