AINEWS
User refusal of unwanted AI features is starting to move product decisions, but only at the edges, and the consent question is still open.
Open a familiar app and a new button has appeared in the feed. Nobody toggled it on. Nobody asked for it.
A recent Gallup poll cited by WIRED found that among Americans 18 to 29, almost half say generative AI is doing more harm than good, the most hostile cohort in the survey. Familiarity with the technology has grown, and so has the anger. What users are asking for, increasingly, is consent: a say in which AI features ship in the software they already pay for or rely on.
In late July, three platforms shipped product changes in response. LinkedIn added a "seems like AI slop" report button for AI-generated content. Snapchat announced that fully AI-generated videos are no longer eligible for its Spotlight discovery feed. Substack added an AI detection tool to police writers. Three different product surfaces, one shared recognition: user refusal is starting to actually move product decisions.
Meredith Broussard put it to WIRED this way: "The AI revolution has happened, and everybody hates it." Her read of the moment is blunt, and it tracks with the data.
Many Americans feel generative features, from data scraping for training to deepfake tools to AI-generated search answers, were pushed into the software they already use with no opt-in. Broussard calls that a consent problem. "Consent is really important," she said. "We can connect that to the way that women, queer people, people of color, and minorities get exploited and abused online. Tech companies have never really been sensitive to issues of consent."
LinkedIn added a report button while continuing to host the AI-assisted posts and writing tools the complaints are about. Snapchat's new rule excludes only fully AI-generated Spotlight videos, leaving AI-assisted or AI-augmented uploads in the feed. Substack's detection tool is itself contested: writers have questioned how it works, what it flags, and whether it will penalize the very human writing it claims to protect. Each of these moves is a real concession. None of them rolls the underlying AI integration back.
A user who has not agreed to have their posts, photos, or writing scraped for model training has no record of refusal. A user who has not opted into AI-generated search answers has no toggle to turn them off. The three refusal surfaces shipped in July are downstream fixes for an upstream problem: most generative features roll out as defaults, with consent extracted after the fact, if at all. The off switch the 18-29 cohort is asking for is not a report button. It is a default-off setting, with an opt-in for the features a user actually wants.
LinkedIn, Snapchat, and Substack do not share a roadmap and do not coordinate product timing. That three of them shipped refusal surfaces in the same month suggests the cost of inaction has started to outweigh the cost of shipping a partial fix. The companies that move first set the template, and the template right now is half-measure: ship an opt-out, not an opt-in. The 2027 product cycle will test whether that template holds or whether a competitor breaks from it.
A 2027 product roadmap that reads the Gallup 18-29 number as edge-case noise will ship another default-on AI feature and discover, on the next quarterly call, that engagement fell and support tickets rose. A roadmap that reads the number as competitive constraint will start designing for an off switch. The difference will show up in the next default-on feature you did not ask for.
Each of the three platforms above has shipped one piece of a refusal surface, not a refusal system. The question is whether the next round of defaults ships an opt-in toggle, or another button you didn't ask for.