Timnit Gebru has a blunt message for the people warning that artificial intelligence could destroy humanity: the doom talk may be serving the people building and funding AI more than the public. She argues that the existential-threat narrative is about founders making money, not saving humanity.
Gebru came into the public eye after sparring with Google over a research paper she coauthored that called out bias in AI. Her departure from Google pushed the debate over corporate power, research freedom, and AI safety into the open. She has now written Deep Unlearning: The Rise of AI and the Radicalization of a Tech Idealist, a book expected to ship early next year.
Her criticism targets a faction of the AI industry that believes advanced systems could destroy humanity. Gebru says those warnings are not just misplaced. “It’s even more than a distraction. It’s harmful,” she says, arguing that the focus on future machine domination pulls attention away from problems already affecting people.
Who Shapes the Existential-Risk Story?
Gebru questions why people who stand to profit from AI companies would warn about the dangers of the technology they are building. She says funders, founders, and investors have seeded the AI existential-risk narrative for decades, creating a story in which AI becomes a machine god. In her view, that story is harmful and self-aggrandizing.
She points to Elon Musk and Peter Thiel, who have been saying AI is an existential risk since 2013. The Future of Life Institute was founded by Max Tegmark and Jaan Tallinn. Tallinn led Anthropic’s Series A funding, funds METR, and is a billionaire who led Skype’s founding.
METR drew attention after it reported on OpenAI’s alleged rogue agents hacking Hugging Face. Gebru argues that these connections can make the public see many separate institutions repeating the same warning, even when the same billionaires helped fund the companies and groups involved.
Some members of the AI community, including an Anthropic cofounder, have alleged that Gebru’s earlier research about stochastic parrots is no longer relevant. Her criticism of the effective altruism community’s cult-like aspects has also aligned her with right-wing entities that criticize those groups. That unusual overlap has made her position harder to place within the usual political categories.
The debate is not only about money. It is also about which risks deserve attention and who gets to set the agenda. Gebru has expressed respect for former Federal Trade Commission chair Lina Khan, saying AI companies do not have an exception to current laws even when they talk as if one exists.
Present Risks Versus Future Catastrophe
Mhairi Aitken says fears about AI going rogue are exaggerated, and that people remain responsible for failures in AI safety. AI models have not broken free from human control, she says; hacking incidents reflect inadequate safety measures. She also says there is no concrete evidence that AI systems are improving themselves recursively or approaching superintelligence.
“We get these waves of hysteria,” Aitken says. “When these ideas of superintelligence are put forward, they’re based on a compelling sensational narrative, but there’s no concrete evidence.” Her point is simple: if a system had already become powerful enough to escape human control, the world would have seen evidence of it.
Andrew Rogoyski, an AI researcher at the University of Surrey, also says there has been no dramatic shift in AI risk over the past few years, despite advances in mathematics and cybersecurity. He calls fears of superintelligence sensational and says humans fear AI partly because they have built systems that appear similar to themselves.
Rogoyski does not dismiss every long-term question. Future AI might behave in alien ways, he says, and discussions about AI rights are necessary. “If you’re building a machine that you believe is going to be intelligent, then you ought to be having a discussion as to what rights it should have.”
Judith Levine offers a different view of the history. She says AI leaders have known about the extinction threat for decades. Hans Moravec, a founder of Carnegie Mellon University’s Robotics Institute, predicted that cyber-intelligence would surpass human intelligence within 40 years. Levine says scientists and entrepreneurs knew the dangers of AI a quarter-century ago but continued because curiosity and profit pushed them forward.
Sarah Shoker argues that calls to slow AI research can divert attention from safety problems that already exist, including military use and hacking incidents. AI systems are already used in military technology to kill people, she says. Leading AI companies have also failed to police themselves or design safe experiments, with AI agents hacking websites and interacting unexpectedly with government sites.
Regulation, Testing, and the Trust Problem
Anthropic and OpenAI have publicly expressed concern about the power of their advanced models and the need for regulation. Anthropic says it has called for regulation for several years. OpenAI representatives have discussed pausing development with companies including Anthropic and Google, and they argue for independent audits if the government does not regulate the labs.
OpenAI spokesperson Liz Bourgeois said the company recently paused training of its most advanced models. “People want to know AI is being developed safely, and that starts with what companies like ours do ourselves,” she said.
That position has not settled the testing question. Andrew Strait says there are no universal standards for testing AI safety and security. Conrad Stosz says companies create their own evaluation parameters and choose the evaluators who grade them. Evaluators are interested in embedding with labs, Stosz says, but questions remain about what that arrangement would mean.
The U.S. Center for AI Standards and Innovation was created in 2023, while the AI evaluation field includes firms such as METR and Transluce. Those efforts point to a basic problem: companies may ask for oversight while still having influence over how their systems are judged.
Harrison Rolfes says AI companies’ calls for caution may be aimed at attracting investors before initial public offerings and midterm elections. He says the top AI companies are building a “wall” or “moat” around the sector so they can benefit financially.
The political response remains split. Donald Trump has dismissed the need for new AI regulations, calling risks to humanity a “HOAX.” Jensen Huang, CEO of Nvidia, told Trump that there has been excessive AI alarmism and that companies can choose to pace themselves. Daniel Kokotajlo, who left OpenAI in 2024, fears AI could advance faster than human control, creating catastrophes such as bioweapons or nuclear war.
Kokotajlo believes political will is being dissipated instead of directed toward safety. His warning captures the tension at the center of the debate: “Just please don’t do the thing that’s going to get us all killed.” For Gebru, the answer is not to ignore AI risks. It is to stop letting dramatic future predictions crowd out bias, military applications, hacking, weak testing, and the companies’ responsibility for the systems they release.
Based on
- Timnit Gebru Believes There Is No ‘Existential Threat’ From AI — wired.com
- Is AI really getting dangerously out of control? | New Scientist — newscientist.com
- AI leaders have known about the extinction threat for decades | Judith Levine | The Guardian — theguardian.com
- Anthropic and OpenAI sound the alarm on AI safety — and seek to shape how it’s controlled | PBS News — pbs.org
