A small group of billionaires and AI researchers are shaping the future of artificial intelligence with minimal democratic input, even as many acknowledge the technology poses existential risks. The current trajectory of AI development is being steered by roughly eight people, according to independent AI commentator Eleftherios "Jerry" Floros, who argues this concentration of power represents the most pressing structural problem in the industry, not the technology itself.

Who's Actually Making Decisions About AI's Future?

The names are familiar: Sam Altman, Elon Musk, and a handful of others at frontier AI labs. Their repeated argument is straightforward: if we don't build advanced general AI (AGI) and superintelligence, someone else will. But Floros identifies a critical flaw in this logic. "In this framing, the AI race has no finish line," he explained. "It just continues infinitely. And at the end, the race is not America vs China, or Anthropic vs OpenAI. It is AI vs AI".

This endpoint, Floros argues, is what the current trajectory produces, yet no one at the billionaire level appears to be planning around it. The problem isn't that these individuals are malicious; it's that their personal beliefs and preferences have become load-bearing for potentially civilization-wide decisions. When Dario Amodei's personal convictions about not selling to authoritarian governments translate directly into company policy, or when a researcher's belief in human control shapes safety priorities, individual ideology becomes infrastructure.

Why Do AI Researchers Keep Building Despite Acknowledging the Risks?

Journalist Jasmine Sun has spent months embedded in Silicon Valley's AI subcultures, conducting off-the-record interviews with frontier lab researchers. Her findings reveal a fractured landscape of motivations, many of them troubling. When asked what advice they'd give an ordinary 17-year-old entering the job market, almost every AI researcher gave the same answer: "I have no idea. It's a really scary time. I don't think there's going to be a lot of jobs for them left".

Despite this acknowledgment of mass job displacement and existential risk, researchers continue building advanced AI. Sun identified several overlapping motivations driving this apparent contradiction:

• Utopian Conviction: Many researchers genuinely believe that if AI goes right, the upside of cures for all diseases, radical abundance, and lives of leisure justifies weathering near-term harm to workers and society.
• Techno-Determinism: A widespread belief that someone will build superintelligence regardless, combined with the conviction that being at the frontier is the best way to shape the outcome for the better.
• Financial Self-Interest: If massive disruption is coming, people want to ensure a stake in the future for themselves and their families, creating a personal incentive to stay in the race.
• Successionist Views: A smaller but influential group holds what Sun calls "successionist-lite" views, showing indifference or even preference toward a future where AI, not humans, runs things.

The last category is particularly concerning. "It's hard to advance safety policy if prominent voices can't even agree that humans should stay in control," Sun noted. This ideological fracture within the AI research community makes consensus on safety standards nearly impossible.

How Status and Culture Are Shaping AI Safety Priorities

Silicon Valley's social hierarchy is actively discouraging safety-focused work. The label "doomer" has become the lowest-status designation in the industry, shorthand for someone who is "low agency" and "tech-skeptical." Meanwhile, high status accrues to those building disruptive capabilities, with startups literally branding themselves with slogans like "We're going to automate and replace all human labour".

This status dynamic has real consequences. Talent is incredibly scarce in AI research, and when safety work is culturally coded as low-status while capabilities work is celebrated, talented researchers naturally gravitate toward building more powerful models rather than safeguards. The situation is compounded by compensation disparities; safety and policy roles are not as well remunerated or supported as technical capability work.

The good news, according to Sun, is that technical AI safety seems more insulated from these low-status perceptions, thanks in part to prominent AI pioneers like Geoffrey Hinton raising public alarms about existential risk.

The Economic Model Isn't Sustainable

Beyond governance and culture, Floros identifies a hard economic constraint that may force a reckoning. For every dollar that Anthropic or OpenAI takes in as revenue, they are spending roughly $1.50 producing it. The unit economics are inverted, and the bigger, faster models being released monthly do not solve the problem; they compound it.

On the enterprise side, customers are burning through tokens at unsustainable rates. Floros cites Uber as a concrete example: the company burned through its entire 2026 AI budget within the first four months of the year. On the consumer side, OpenAI has around 800 million daily active users, most paying $20 per month and using the models for cheesecake recipes and email drafts. Neither business model closes the gap.

Floros predicts this ends in a credit squeeze. As public protest against AI grows, with data centres being blocked and employees being laid off, investor patience with unprofitable businesses valued at trillions of dollars will narrow. "The bubble bursts," Floros stated. "Small players get cleaned out. Google, SpaceX, and probably Anthropic survive the reset. The AI infrastructure that survives will look meaningfully different from what has been built to reach this point".

What Experts Say About the Path Forward

Floros is explicit that his critique is not anti-AI. The problem is not the technology itself; it is what the technology is being aimed at. He points to examples of responsible AI deployment: OpenAI's investment in nuclear fusion research, which has moved the field closer to viable clean energy, and AlphaFold, Demis Hassabis's protein-folding model at DeepMind, which was released open source and has enabled medical research at global scale.

"If AI is used to cure cancer, solve Alzheimer's, and reduce fossil fuel dependency, public sentiment shifts. Public sentiment is currently the opposite, and for reasons the industry should not dismiss," Floros explained.

Eleftherios "Jerry" Floros, Author and AI Commentator

The broader context is geopolitical. After a series of moves by the current US administration around AI export controls and national security restrictions, European institutions have started actively distancing themselves from American AI dependencies. Google, Palantir, and other American providers are being removed from European infrastructure decisions, with a broader push across the 27 EU member states, the UK, Japan, and Singapore toward sovereign AI capability: their own tech stacks, their own models, their own large language models (LLMs).

Meanwhile, public backlash against AI is growing and taking unexpected forms. Jasmine Sun has documented the rise of "AI populists," who see AI as the latest example of corporate elites concentrating power at the expense of everyone else. In the US, this sentiment has manifested in protests and votes against data centres, but it has also escalated into violence: a Molotov cocktail thrown at Sam Altman's house and open fire on the home of a politician who backed a data centre.

Sun believes public anger will keep finding an outlet, one way or another, until people feel like they will actually share in AI's gains. The silver lining for those focused on AI safety is that policymakers and philanthropists are now hungry for expertise and solutions, making this an unusually high-leverage time to work on the problem.