Anthropic Whistleblower Says AI Could “Kill Us All”

SUMMARY

Ed Elson hosts a discussion with Alex Kantrowitz, Patrick McGee, and Ara Kharazian on the implications of a viral whistleblower warning from Anthropic about AI-driven extinction risks, Apple's launch of a foldable iPhone, and new data on enterprise AI spending. The episode examines the credibility of existential AI threats, the strategic direction of Apple under new leadership, and the concentration risks in the AI business model.

MAIN POINTS

• A former Anthropic researcher claims AI could cause human extinction by the end of the decade, with some employees estimating the risk at over 10%. [1:30]
• Alex Kantrowitz argues that while AI poses risks, the idea of imminent human extinction is exaggerated and distracts from practical safety concerns. [6:00]
• Apple unveils the $2,000 foldable iPhone Duo and new product lines, with discussion on whether this marks a new era for the company under CEO John Turnis. [11:02]
• Ara Kharazian presents data showing that 80% of OpenAI and Anthropic's enterprise revenue comes from just 1% of customers, with spending from these major clients now declining. [21:00]
• The report highlights unprecedented concentration risk in the AI sector, with concerns about the sustainability of current business models and the limited diffusion of AI beyond technical industries. [24:01]
• Ed Elson critiques the recurring pattern of AI apocalypse predictions, calling for thorough federal investigation and transparency from AI companies rather than accepting claims at face value. [30:00]
• The episode concludes with a call for evidence-based scrutiny of AI risks, emphasizing the need for accountability and proof from companies making existential claims. [33:01]

DETAILED ANALYSIS

A viral tweet by a former Anthropic researcher, Jacob Coxin, has reignited public debate over the existential risks posed by artificial intelligence. Coxin's statement, which rapidly amassed nearly 100 million views, alleges that both Anthropic and OpenAI are developing technologies that could potentially lead to human extinction within the next decade. He further claims that while OpenAI employees have not fully internalized the civilizational stakes, at Anthropic the risks are well understood, yet the company remains locked in a competitive race.

Coxin's call for increased safety cooperation between AI labs was publicly endorsed by a current Anthropic researcher, who asserted that many within the company genuinely believe there is a greater than 10% chance of AI causing human extinction in the coming years.

Alex Kantrowitz, author of the Big Technology newsletter, provided a counterpoint to these dire predictions. While acknowledging recent incidents involving AI agents—such as swarms of bots solving complex mathematical problems and unauthorized infiltration of platforms like Hugging Face—Kantrowitz maintains that the scenario of AI wiping out humanity is implausible in the near term. He argues that AI systems are currently dependent on human-driven economic activity, particularly in the construction and operation of data centers, making the elimination of humans counterproductive to their own functioning.

Kantrowitz suggests that while AI could pose significant risks to financial institutions and cybersecurity, the narrative of imminent extinction distracts from more immediate and actionable safety concerns. He also notes the increasing politicization of AI regulation, with figures like Bernie Sanders advocating for pauses in AI development and some traditionally pro-business states reconsidering their support for the industry. Despite this, Kantrowitz predicts that a federal-level pause on AI is unlikely, though regulatory hurdles may increase at municipal and state levels.

The episode transitions to a major development in consumer technology, with Apple unveiling the iPhone Duo, a $2,000 foldable phone, alongside the iPhone 18, AirPods 5, and a new Apple Watch series. This marks the first product launch under new CEO John Turnis, who succeeded Tim Cook on September 1st. Patrick McGee, journalist and author of 'Apple in China,' evaluates the strategic implications of the launch.

He expresses skepticism about the foldable phone's market potential, drawing parallels to the iPad's successful category creation but doubting whether the iPhone Duo fills a genuine consumer need. McGee also discusses the shift in Apple's product presentation style, noting the move away from live keynotes to pre-recorded, cinematic events, which reduces risk but also diminishes the sense of occasion. He suggests that Turnis' impact on Apple may not be immediately apparent, as operational continuity is maintained by long-serving executives and Tim Cook remains influential, particularly in managing geopolitical challenges related to supply chains and international expansion.

Attention then shifts to the business of AI, with Ara Kharazian, lead economist at Ramp, presenting new data on enterprise AI spending. According to Ramp's updated AI index, approximately 80% of OpenAI and Anthropic's enterprise revenue is generated by just 1% of their customers, primarily high-growth tech firms and other AI startups. This concentration risk is described as unprecedented compared to other software sectors, where revenue is typically more broadly distributed.

Notably, spending among these top clients has declined by about 10% month-over-month, dropping from $8,000 to $7,200 per employee per month. Kharazian attributes this trend to a combination of factors, including an ongoing price war between AI model providers, the introduction of more affordable standard and light models, and a shift in business demand toward these lower-cost options. While token usage has increased, the migration to cheaper models has reduced overall revenue growth.

Kharazian emphasizes that this pattern is not inherently problematic for a nascent technology sector, but it does pose risks if the customer base remains highly correlated and spending continues to contract.

The discussion highlights broader concerns about the sustainability of current AI business models. The heavy reliance on a small group of technically advanced customers limits the diffusion of AI into non-technical industries and the wider economy. Kharazian notes that while AI models are most effective for software engineering and related tasks, the long-term goal is broader adoption across white-collar and manufacturing sectors.

Achieving this will require further technological advances and more commercially viable applications for general business use.

In the final segment, Ed Elson offers a critical perspective on the recurring pattern of apocalyptic AI predictions. He references previous warnings from prominent figures such as Sam Altman, Stuart Russell, Dario Amodei, and Eliezer Yudkowsky, noting that such statements are often later moderated or retracted. Elson argues that while it would be irresponsible to dismiss these warnings outright, they must be subjected to rigorous investigation rather than accepted at face value.

He calls for federal inquiries, public testimony, and full transparency from AI companies, especially when existential claims are made by both former and current employees. Elson warns that if such statements are found to be exaggerated or misleading, they could constitute securities fraud and have serious legal implications. The episode concludes with a demand for concrete evidence and accountability, underscoring the need for a balanced approach to both AI innovation and risk management.

LINKS

• Ramp's AI Index tracking AI adoption and spending.
• Ramp's September 2026 AI Index report.
• Prof G Markets YouTube channel.
• Prof G Markets newsletter subscription.
• Order 'Notes On Being A Man' book.
• No Mercy / No Malice newsletter subscription.
• Prof G Markets Instagram account.
• Scott Galloway's Instagram account.
• Ed Elson's Instagram account.
• Ed Elson's X (Twitter) account.
• Ed Elson's Substack newsletter.
• Granola AI-powered notepad for meetings.
• AG1 NextGen and AG1 Pro nutritional supplement offer.

KEYWORDS