Knock, Knock, Who’s there?
Jennifer DeStefano’s Senate testimony is not written like any tech policy document. It is a mother trying to explain how her body reacted before her mind could catch up. Picking up the phone, she describes hearing her teenage daughter sobbing and calling “Mom.” Then a man’s voice comes in and demands money. DeStefano writes that she could not treat it as a scam. It was not only the voice, but the crying, the inflection, the things she thought were uniquely her child fueled her fear. Only later does she learn her daughter is safe, and what she trusted was a clone.
I keep coming back to that detail about inflection since it shows why voice cloning is more than just another chapter in misinformation. We have learned to be suspicious of links, and many of us become scrupulous about images. Yet a voice carries a different sense of intimacy. It is the feeling that what you are hearing is an authentic individual rather than some random performance. That is exactly what makes it vulnerable against criminals and useful to industries seeking frictionless authentication. It is also why it may be politically consequential when voices become cheap to manufacture.
Verma’s reporting on AI voice scams raises a similar point in a different register. The story is not that people are gullible, but how scammers are using AI to fake a familiar relationship during one phone call, so quickly that the usual checks cannot happen, and victims lose thousands before they can sit back and think. The FTC warning is even more blunt. A scammer can clone a loved one’s voice with a short clip pulled from online content and a voice cloning program, and then use that to stage an emergency that sounds and feels real. The technology does not create fear; it makes fear more authentic.
This is where I find Jasanoff’s The Ethics of Invention clarifying, especially her push against the comforting idea that harms are simply “unintended consequences” that nobody could foresee. Julián Goñi’s review of her book captures the argument cleanly: Harm can occur without apparent intention because, in many technological systems, “no single actor is ever in charge of the entire big picture.” In other words, the problem is not only about tools with side effects, but about how responsibility is structurally impossible to locate, making it easy to offload, delay, or deny. DeStefano’s call is frightening precisely because it is hard to pinpoint one single decision-maker who caused it, even though it is obviously made possible by many decisions. What Jasanoff helps me notice here is not the absence of accountability, but its diffusion across a system in which no single actor can be held responsible for the whole.
Get Stnzlip’s stories in your inbox
Join Medium for free to get updates from this writer.
While some may argue that calling voice cloning an emerging technology flatters its novelty, it instead explains why governance keeps lagging. Rotolo et al argue that emerging technologies tend to be fast-growing and impactful, but also highly uncertain and ambiguous, including ambiguity about what the technology is for and how it will be used. Voice cloning fits that pattern. It is being packaged as creativity and productivity, and simultaneously being weaponized for fraud and coercion. The same feature that makes it attractive, a convincing human signal produced cheaply, also makes it hard to constrain.
Collingridge’s dilemma of control describes the trap that follows. Early in a technology’s life, it is easier to change direction, but the harms and pathways are unclear. Later when harms become visible, the technology is then already embedded in markets and routines, and steering is now costly and politically contested. Voice cloning is moving through that path fast. It is no longer a lab demo; it is becoming a feature, an option in a menu, an add-on for customer service, content production, sometimes even identity verification. The more normal it becomes, the more any restriction can be framed as anti-innovation, even if the restriction is simply a way of protecting ordinary people from being forced into constant verification work.
With the regulation of voice cloning still ambiguous, the advice circulates after scams appear to hold individuals responsible and put pressure on the public. The Guardian reported warnings in the UK about social media videos being exploited for voice cloning; however, it also included advice from a bank security officer to use a pre-agreed safe phrase with family members. This is sensible advice, and I do not want to mock it. Yet I also think it should make us uneasy. A safe phrase is a grassroots patch for a broken trust environment. It is the kind of patch that quietly changes what counts as a normal relationship and trains people to treat everything, even a familiar voice, as a suspect unless proven otherwise.
That is why I think the proposed solution is a suspicion tax. It is the extra time, cognitive effort, and emotional regulation required to authenticate everyday interactions that used to be self-evident. Some people can pay that extra tax more easily than others. They are individuals with flexible jobs, strong networks, and multiple communication channels open for pause and verification. However, isolated, overworked, or emotionally exhausted individuals often cannot. Furthermore, within organizations, power and hierarchies shape who feels permitted to challenge a request. A junior employee may not dare to interrogate what appears to be a superior’s call, especially in cultures where efficiency and deference are rewarded.
So here I am with a question, less of a cybersecurity one but more of a political one. If voices stop being proof, what kind of society are we building without that crucial shortcut? One path treats authenticity as a commodity. You buy it through more surveillance, more gatekeeping, more individual verification, and more burdens placed on people with the least time to comply. Another path treats trust as public infrastructure, something that must be governed top-down, with responsibility pushed up, and with attention to unequal vulnerability. Jasanoff’s call for “technologies of humility” is relevant here, asking institutions to take uncertainty and vulnerability seriously, and to govern with public participation rather than assuming experts and markets will automatically deliver legitimacy.
DeStefano’s testimony ends as a warning, but it also reads to me as a sociological clue. The scam worked since a cloned voice could easily temporarily occupy the place where trust usually lives. The hard part is not learning to distrust that place. It is deciding who should pay for the repair and refurbishment.
