Even when we look, we can miss what’s right under our noses. The well-known “invisible gorilla” effect illustrates a broader truth: human perception has systematic blind spots that extend across the multisensory information we process every day. As we develop super-intelligent AI, understanding these human limits of attention becomes critical. We must be able to predict and measure when human lapses in perception can be exploited—allowing important information to pass unnoticed. Accounting for human perceptual blind spots is essential to building AI that is safe, trustworthy, and aligned.