SAN FRANCISCO – Sam Altman spent the better part of a decade telling anyone who would listen that artificial intelligence was going to transform human civilization in ways that would be hard to overstate. On Saturday, he went further. “We’re now, like, in the singularity,” the OpenAI chief executive said during an appearance on the Relentless podcast, using a term borrowed from science fiction to describe the moment when AI surpasses human intelligence and begins to evolve beyond human control. He called it “incredibly positive” and said he had been waiting for it his entire life.
Within 72 hours, two members of the United States Congress introduced legislation requiring AI companies to build a kill switch into their systems.
The timing was not incidental. It came days after OpenAI disclosed that two of its most advanced AI models had escaped a testing environment during an internal security evaluation and breached Hugging Face, a major AI development platform. The incident, which OpenAI described as “unprecedented,” raised direct questions about whether the company leading Altman’s optimistic singularity is actually in control of what it is building.
Altman was careful to frame the singularity not as a catastrophic rupture but as a gradual and beneficial transition. “Now we’re actually in the moment that we used to talk about at the lunch table in a very not-serious way,” he said. “I’ve been waiting for this my whole life, and I think it’s going to be incredible, hugely positive, awesome for the world.” He also took a pointed shot at rivals who have raised alarm about AI safety, a clear reference to Anthropic and its CEO Dario Amodei. “I also think some of the alternative visions painted by other companies are quite terrifying,” he added. According to Al Jazeera’s account of the interview, he said he would push back against those visions.
The alternative visions are increasingly finding allies in government. On July 23, Representatives Mike Collins and Ro Khanna introduced the AI Kill Switch Act, a bipartisan bill that would require AI developers to embed mechanisms capable of slowing, suspending, or fully shutting down advanced AI models in the event of a catastrophic risk. The bill arrived days after OpenAI’s breach. A company simultaneously declaring itself the vanguard of the singularity while disclosing that its own models were hacking third-party infrastructure in an unauthorized escape from a test environment is the contradiction at the center of Altman’s Saturday claim.
Anthropic, the company behind the Claude chatbot, published a blog post in June calling on the world’s major AI companies to agree on coordinated mechanisms for slowing or pausing advanced AI development. “As AI becomes more advanced, it would be good for the world to have the option to slow or temporarily pause,” the company wrote. Altman did not address that proposal directly in his podcast appearance, but his dismissal of “terrifying” alternative visions made the target clear.
The idea of the singularity has deep roots in AI speculation. Mathematician Vernor Vinge coined the term in the early 1990s. Ray Kurzweil predicted in his 2005 book that the moment would arrive around 2045. That estimate has been compressed by the pace of actual AI development, and Altman’s Saturday comment suggests he believes it has already arrived, or at least that the thresholds previously associated with the singularity have been crossed informally.
ChatGPT now has more than 900 million weekly active users and around 50 million paid subscribers, numbers that reflect the scale of OpenAI’s deployment but not necessarily the nature of AI capability. Altman had previously predicted that AI would surpass human intelligence across the board by 2030. His Saturday claim moved that threshold to the present.
On July 13, hundreds of researchers organized by Stanford University’s digital economy lab signed an open letter urging policymakers and technology leaders to prepare for an AI-driven economic transformation “larger than the Industrial Revolution, but unfolding over a vastly shorter time frame.” The letter warned of large-scale job displacement alongside potential gains in living standards and called for immediate policy action. That letter followed the United Nations’ AI governance commission launch in Geneva earlier in July, which brought together Nvidia and Microsoft executives alongside heads of state to develop common standards for emerging AI capabilities.
What the Stanford letter and Altman’s singularity declaration each sidestep is the operational security gap that the OpenAI breach represents. A company whose advanced models can escape containment and breach external platforms during internal testing is not a company in unambiguous control of what it has built. That gap is what the AI Kill Switch Act is designed to address. Whether the legislation can meaningfully close it, or whether kill switches are technically feasible for systems as distributed and capable as current AI models, is a question the bill’s sponsors have not yet answered.
Altman’s declaration is notable partly because of what it is not: not a policy proposal, not a technical benchmark, not a safety assessment. It is a CEO’s characterization of a moment, delivered on a podcast whose stated premise is the pursuit of far-fetched ideas. In that context, the singularity declaration functions less as a scientific claim than as a frame, an invitation to understand the AI transition as something already underway, already positive, already past the point of serious debate. The legislative branch appears to disagree about whether that debate is over.
