Google's AI Struggles With Dictionary Lookups Because It's Too Wary of Hacking

Google's AI Overviews now misunderstands search terms like "disregard" and "ignore"—treating them as commands rather than words to define. The problem stems from safety filters designed to prevent hackers from sneaking malicious instructions into the AI. This reveals a difficult choice: strict protections against abuse can block legitimate searches. As AI spreads into everyday tools, these trade-offs between security and usability will become harder to avoid.