Whenever I examined the workings of a public institution, I found that errors of design were more dangerous than errors of calculation. In this case the design couples a sweeping law about online harm with a very specific instrument, a facial scan inside an entertainment app owned by one of the largest firms on earth. The legislators point to twenty twenty five deadlines and fines measured as a share of global revenue, and Twitch dutifully produces a mechanism that can count and classify faces at scale. Yet my experience with machinery teaches me to ask where the cogs will end up once the immediate political pressure has passed. Will the same apparatus be quietly extended from mature streams to political content, to protests, to anything that makes a minister uneasy? Will the data that today are said to remain on the user's device tomorrow be mirrored for fraud detection, then advertising, then some vague notion of trust and safety? A capricious algorithm hidden in the cloud is worse than a badly printed timetable, because the citizen cannot see where the error arises or demand that it be corrected. If Parliament wished to protect children, it might have invested equal effort in education, parental controls and simpler, non biometric proof systems, rather than pushing every platform toward the same slightly magical solution and hoping it will do the moral work for them.