The global grassroots gatherings trying to humanize the AI boom

Twenty strangers sat in a circle in an empty San Francisco office on a recent Saturday, talking not about funding rounds or model benchmarks but about the messy business of sharing neuroscience data. A high schooler was there. So was a neuroscientist, a few startup founders, some postdocs. This was an AI Salon, one of roughly 40 such gatherings that neuroscientist and AI safety researcher Ian Eisenberg has hosted since starting the group shortly after ChatGPT's launch in 2023.

The premise is almost quaint by Silicon Valley standards: no pitch decks, no hackathon energy, just people from wildly different backgrounds forced to keep an open mind and speak at least once. Eisenberg's group now has chapters in Lagos and Bengaluru, run entirely by volunteers, capped at about 20 people per session. It borrows its name and format from 17th- and 18th-century European salons, the ones where Enlightenment ideas got hashed out over wine in wealthy people's living rooms. The irony that today's version is happening in the shadow of Silicon Valley's biggest tech companies is not lost on organizers.

A bigger, more corporate cousin exists in the AI Collective, which has swelled to more than 200,000 members across 200 chapters in 50 countries since its 2023 founding. Chief operating officer Catherine McMillan says the goal is deliberately local: build community around AI rather than let a handful of American firms dictate the terms of the conversation. In Nairobi, where AI adoption sits under 10%, organizer Francis Kangure frames his 3,000-member chapter as a bridge between African talent and global opportunity, at a moment when data privacy and job losses worry as many people as the technology's promise excites them. In Bengaluru, chapter head Abhishek Kumar sees a different gap: India is Anthropic's second-largest Claude market, yet he worries the country risks becoming a permanent consumer of foreign-built models rather than a producer of its own, leaving non-English speakers and non-metro residents further behind.

But the uncomfortable question, raised bluntly by Utrecht University professor Payal Arora, is whether any of this actually reaches the people AI is most likely to disrupt. Members of these salons and collectives tend to already work in tech, she points out, meaning the gatherings risk simply reproducing the same elite networks and power imbalances they claim to challenge. It takes real social and cultural capital just to know these meetings exist, let alone to feel comfortable walking into one. Even Satyanarayana, who organizes the Bengaluru salon, admits as much: it's a small subset of people, not remotely representative of the population AI will affect most.

Meanwhile a rougher, angrier strand of organizing is growing alongside the salons. Groups like Pause AI and Stop AI have moved from Signal group chats to sidewalk protests outside OpenAI, Anthropic, and Google DeepMind, demanding slower development, real safeguards, or a bigger cut of the profits for everyone else. Compared to that, the calm, RSVP-only conversation happening in a San Francisco office can look almost genteel. Both approaches share a starting assumption, though: nobody is waiting for governments to figure this out for them.