Researchers claim breakthrough in fight against AI’s frustrating security hole

Prompt injections are the Achilles’ heel of AI assistants. Google offers a potential fix.