What happened
After public outcry, Google rolled back its generative-AI satellite-image tool on Google Earth nearly immediately, and Meta disabled Instagram's AI deepfake feature after three days of criticism and viral videos. LinkedIn added a "seems like AI slop" button, Snapchat banned fully AI-generated videos from its discovery feed, and Substack deployed AI detection. Companies including McDonald's and Coca-Cola faced online backlash for using generative AI in advertising.
Why it matters
A Gallup poll shows Americans' familiarity with generative AI now coincides with negative attitudes, especially among 18–29 year-olds where almost half view it as doing more harm than good. The core complaint is lack of consent: users were not asked before their data was scraped for model training, before deepfakes were enabled on their accounts, or before AI answers flooded search results. Consent issues particularly affect women, queer people, and people of color who have historically been exploited online.
What to watch
Public pressure—both online and through data-center protests uniting Americans across the political spectrum—has proven effective at forcing reversals. Software developers report being forced to use large language models despite doubts about their utility and ethics, signaling tension ahead in workplace adoption.
Summaries like this, in your inbox every morning.
The recent rollbacks reveal a shift in how tech companies respond to public pressure around generative AI. For years, the industry rolled out features rapidly with minimal consideration for user preferences or ethical implications. The pattern described by Nick Seaver, an associate professor of anthropology at Tufts University, captures this dynamic: "They just roll them out and see what sticks, because nobody really knows what this is for." What has changed is that coordinated online backlash—amplified by viral videos, reporting, and cross-partisan organizing around data centers—now moves the needle fast enough to force reversals within days or weeks rather than allowing features to become entrenched.
The consent problem runs deeper than individual feature complaints. Users were not asked before their online interactions were harvested for AI training, before Instagram enabled deepfakes on their accounts, or before Google populated search results with AI-generated answers. This lack of agency reflects a broader pattern in tech: features roll out top-down, and users are expected to adapt or opt out through often opaque settings. For workers, the complaint is equally stark—Block employees and others are being mandated to use large language models despite believing the tools are not fit for purpose and carry ethical risks. The backlash appears to be rebalancing that dynamic by making companies accountable to user and employee sentiment in real time.
For example, today's edition would include:
AI-summarized, only the topics you pick: one digest a day via Email, LINE, or Slack.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
