Venture capitalist and All In Podcast co-host Jason Calacanis joined Adam Curry on episode 1905 of the No Agenda Show to dismantle what he sees as a dangerous fusion of messianic AI safety rhetoric and old-fashioned regulatory capture. The conversation, published September 20, 2026, offers a rare window into how Silicon Valley's elite are framing the AI debate ahead of the 2026 midterms, and why the stakes extend far beyond technical questions of model alignment.
Calacanis reserved his sharpest criticism for Anthropic, the AI lab founded by former OpenAI researchers and backed by billions in venture capital. 'They actually think that they are creating a god. They're not. They're creating software,' he told Curry, comparing the company's effective-altruism roots to the doomed engineers in the film Prometheus. Curry played clips of University of Washington professor Pedro Domingos describing an Anthropic 'funeral' for a retired Claude model, underscoring what Calacanis characterizes as a quasi-religious culture inside the company. The implication is stark: if the leading voices in AI safety are driven by metaphysical beliefs rather than engineering pragmatism, their policy prescriptions deserve far more scrutiny than they typically receive.
But Calacanis argues the real story is not the theology—it is the money and power that theology justifies. Frontier labs, he says, are lobbying Washington to impose safety rules that would effectively gate open-source competitors like DeepSeek, GLM, and NeMoTron. In other words, the same companies warning about existential risk are simultaneously working to entrench their own market dominance. That dynamic matters enormously for the future of AI innovation, because open-source models have become the primary check on the pricing power and political influence of a handful of well-capitalized labs.
The episode also touches on Microsoft AI CEO Mustafa Suleyman's new code of conduct and his public criticism of Anthropic's 'model welfare' doctrine, a reported security incident at Hugging Face, and calls for an AI 'kill switch.' Each of these stories, Calacanis suggests, is being weaponized to justify preemptive regulation that would freeze the current competitive landscape in place. For listeners unfamiliar with the No Agenda Show, the program is a long-running, listener-supported podcast known for its 'media deconstruction' approach and value-for-value funding model, available at https://www.noagendashow.net/.
The political dimension is equally important. Curry and Calacanis discuss the weaponization of data centers by candidates like Texas Democrat James Talarico, Trump's revocation of press credentials for CNN, MSNOW, and Politico, and rising diesel prices tied to Ukrainian strikes on Russian refineries. These are not separate stories; they are symptoms of a broader information war in which AI, energy, and media access are all contested terrain. Calacanis also shares backstage details from the All In Summit, including Jensen Huang fumbling a speakerphone call from President Trump and a tense exchange with Meta's Dina about a record $17 billion settlement during a data-center segment.
For investors, founders, and policymakers, the takeaway is clear: the AI safety debate is no longer just about technology. It is about who gets to build, who gets to compete, and who gets to decide the rules. If Calacanis is right that Anthropic and its peers are engaged in regulatory capture dressed up as existential risk management, then the 2026 midterms could determine whether open-source AI survives as a meaningful alternative—or becomes a footnote in a cautionary tale about concentrated power.
