Humanoid robots can fail and do unsafe things in countless ways – hardware malfunctions, software glitches, cyberattacks, freak edge cases, or human operator error. A hundred years ago, there were 20 vehicle accident deaths per 100 million miles in the US. Today, that number is under 1.5 – but not zero. Robots won’t be 100% safe – and maybe that shouldn’t be the expectation. We're in the “early automobile” phase for humanoids. We don’t yet have enough real-world data on robots working among people to define what “safe enough” truly means. But robots do need to be among people – to assist the elderly, and to take on dangerous, repetitive, and tedious tasks. We’ll need to establish reasonable safety standards that protect people without stalling progress. If we’re too risk-averse, we risk stagnation; if we’re too cavalier, we invite harm.