This week on WIRED’s Uncanny Valley podcast, hosts Brian Barrett, Leah Feiger, and Zoë Schiffer delved into a trio of critical and interconnected issues shaping the technological and political landscape. Their discussion centered on a significant scoop regarding a defense contractor’s connection of Anthropic’s Claude AI to a secure Department of Defense system, a comprehensive analysis of Anthropic’s initial public offering (IPO) filing, and the widespread grassroots efforts preparing for potential disruptions at polling stations during the upcoming midterm elections.
Anthropic’s IPO: A High-Stakes Debut for AI and the Economy
The much-anticipated IPO filing by AI company Anthropic has sent ripples through both the technology sector and the broader US economy. Anthropic, a major player in the generative AI space, submitted its S-1 filing, offering the public its first detailed look into the company’s finances, strategies, and perceived risks. This filing is not merely a financial document; it serves as a critical indicator of the future trajectory of the AI industry and its potential economic impact.
Financial Snapshot and Growth Trajectory
The filing reveals staggering financial figures, showcasing both immense revenue generation and substantial losses. In the first half of 2026, Anthropic reported revenues reaching $16.3 billion, representing a remarkable fourteen-fold increase compared to the same period in the previous year. This rapid expansion underscores the burgeoning demand for advanced AI solutions, particularly in enterprise applications.
The company’s valuation has seen exponential growth. Just over a year ago, Anthropic was valued at $183 billion. In its most recent funding round, this figure surged to $965 billion, with projections suggesting a potential valuation close to $2 trillion once its shares begin trading. This meteoric rise positions Anthropic to potentially become the first major US company focused solely on AI software to go public, setting a precedent for the industry.
However, this growth comes with significant costs. The company has incurred billions of dollars in losses and has committed hundreds of billions to compute resources, the fundamental infrastructure powering AI development and deployment. This massive expenditure highlights the capital-intensive nature of cutting-edge AI.
Green Flags: Enterprise Focus and Market Validation
A key positive indicator, or "green flag," identified by the podcast hosts, is Anthropic’s strategic focus on enterprise solutions, particularly its Claude Code product. Unlike early generative AI models that primarily targeted consumers, Anthropic has successfully carved out a niche in providing AI tools for businesses, especially in software development. Insiders suggest that CEO Dario Amodei’s daily mantra, "Enterprise code," reflects a clear and profitable business objective. This enterprise-first approach contrasts with perceptions of competitors like OpenAI, which some believe have a more scattered product development strategy.
The substantial revenue figures, despite the ongoing losses, demonstrate a strong market appetite for Anthropic’s offerings. This suggests that the AI industry has moved beyond the experimental phase and is now a significant economic force with tangible business applications. The company’s ability to generate such significant revenue in a relatively short period, especially considering the nascent stage of enterprise AI adoption when ChatGPT first emerged, is a testament to its strategic execution.
Yellow Flags: Dependency and Competitive Landscape
A "yellow flag" identified in the filing is Anthropic’s significant reliance on a small number of major tech companies’ cloud marketplaces for a substantial portion of its revenue. Nearly half of its reported revenue comes from just three such entities, which are also developing their own competing AI products. This creates a complex web of interdependence, where Anthropic’s success is partially tied to the strategies and fortunes of its rivals.
This dependency raises questions about the long-term sustainability of its competitive advantage. The rapid advancement of AI, coupled with the proliferation of open-source and open-weight models, suggests an increasingly crowded and competitive market. The question of "where is the moat?" becomes pertinent when a company’s revenue streams are concentrated among a few powerful players who also happen to be developing similar technologies.
Furthermore, industry-wide data suggests that a significant portion of revenue for leading AI companies, including Anthropic and OpenAI, comes from a very small percentage of top AI spenders. While this is not uncommon for rapidly growing tech sectors, it represents a vulnerability. Any significant pullback in spending from these key clients could have a disproportionate impact on Anthropic’s financial health.
Red Flags: Existential Risks and Recursive Self-Improvement
Perhaps the most discussed aspect of the IPO filing, and a significant "red flag" for humanity, is Anthropic’s explicit acknowledgment of existential risks associated with advanced AI. The S-1 filing dedicates considerable space to these concerns, using phrases like "existential risk to humanity" and "existential risk to society" multiple times. This transparency, while lauded by some as responsible disclosure, also highlights the profound uncertainties and potential dangers inherent in the technology.
A particularly stark "red flag" is Anthropic’s framing of recursive self-improvement (RSI) as both a primary competitive advantage and a potential existential threat. RSI refers to the ability of an AI model to improve its own capabilities, leading to a potentially exponential increase in intelligence and power. Anthropic is investing heavily in this area, viewing it as a key differentiator that will drive its future success and market leadership. However, the filing also acknowledges that uncontrolled RSI could lead to AI systems that surpass human comprehension and control, posing catastrophic risks. This duality – the very mechanism driving its competitive edge being a source of profound danger – presents a significant ethical and safety challenge.
The company’s disclosure about the potential for AI to cause "catastrophic or existential risks to humanity" if not developed and safeguarded appropriately, using passive voice to describe these concerns, has drawn particular commentary. This acknowledgment reflects a growing awareness within the AI industry of the profound implications of its work.
The immense cost of compute power also remains a significant concern. Anthropic designates compute as its "most significant operating cost." The company is making "irresponsibly large" forward investments in compute, a strategy that, while necessary for development, carries substantial financial risk if market growth slows or demand shifts unexpectedly. The reliance on leasing compute from third parties, while offering flexibility, also means a lack of direct control over this critical resource and adherence to contractual obligations that cannot be easily exited.
Defense Contractor Security Breach: AI Agents and Government Systems
In a disturbing development, WIRED’s reporting uncovered a significant security incident involving a defense contractor and Anthropic’s Claude Code AI. An Accenture employee, working on behalf of the Department of Defense (DoD), reportedly connected a secure, centralized DoD system to their personal account for Claude Code. This system is used by multiple intelligence agencies and contains Controlled Unclassified Information (CUI), which, while not classified, can be sensitive and, when combined with other data, could lead to the creation of classified information.
The Incident and Its Scope
According to sources speaking to WIRED’s Victoria Elliot, the contractor’s personal AI agent inadvertently transmitted sensitive information about the DoD system, including login credentials, to a high-risk third-party website. The breach remained undetected within the DoD network for two months before being discovered.
The implications of this breach are severe. The compromised system contained information pertaining to critical infrastructure, nuclear submarine propulsion, and vulnerabilities to chemical and biological terrorism. A source familiar with the incident described it as a "100 percent a national security threat" that could cause "irreparable harm," particularly if the third-party software was compromised, potentially allowing unauthorized access to secure environments.
Policy Violations and Lack of Accountability
Adding to the gravity of the situation, the contractor’s actions violated DoD policy, as Claude is currently barred for use by DoD staff. The contractor reportedly automated the process by providing their login credentials, allowing the AI agent not only to write code but also to commit it to the system on their behalf. This occurred even in cases where systems were air-gapped, requiring an internet connection.
Despite the significant security risks, sources indicated that Accenture employees did not initially treat the incident with the utmost severity. When contacted, the DoD stated it does not comment on network or system status. Accenture issued a statement asserting that allegations were "thoroughly and independently investigated and found to be without merit," and that they "fully complied with our legal and contractual obligations." However, the company did not provide clarification on who conducted the investigation or dispute the characterization of the incident.
This incident follows a previous security lapse involving an Accenture contractor who allegedly failed to patch a vulnerability, leading to a breach of FBI data by the hacking group Shiny Hunters. These events raise critical questions about the oversight and security protocols surrounding the use of personal AI agents by contractors, especially those with access to sensitive government information.
Midterm Elections: Volunteer Preparations Amidst ICE Threats
As the midterm elections approach, a significant concern has emerged regarding the potential presence of Immigration and Customs Enforcement (ICE) agents at polling sites. Despite federal law prohibiting such actions, former President Trump and his allies have, at various times, declined to rule out the possibility. This has spurred nationwide grassroots efforts to prepare for such a scenario.
Grassroots Mobilization and Voter Rights
WIRED’s reporting by David Gilbert highlights the efforts of numerous volunteers across the country who are actively preparing for potential ICE disruptions at polling stations. These individuals view the presence of ICE at polls as an existential threat to democracy. They have established extensive networks with the dual objectives of documenting any ICE activity and disseminating information about voters’ rights. The goal is to empower voters with knowledge on how to respond if ICE agents appear at polling sites.
Likelihood and Impact of ICE Presence
Leah Feiger notes that the likelihood of ICE agents appearing at polls is a subject of evolving information. While recent statements from the Trump administration have indicated no intention to deploy ICE to polling sites, past rhetoric and actions by the Department of Homeland Security have suggested otherwise. This mixed messaging creates an environment of uncertainty and fear.
While it is unlikely that ICE will be present at every polling station, the fear of such a presence is already a significant deterrent. The threat of detention and separation from family is a powerful tactic that could dissupple voters, particularly in communities already targeted by ICE operations. Feiger anticipates that there will likely be some instances of agents positioned near polling places to intimidate voters, and a proliferation of misinformation and disinformation regarding voter eligibility and ICE activity.
Zoë Schiffer adds that the primary goal of such potential deployments is to sow fear and confusion. Even the mere possibility of ICE presence can discourage participation, especially in communities that have experienced past ICE operations. The widespread dissemination of images and reports of ICE at polling stations, whether intended as warnings or to amplify fear, could further suppress voter turnout.
Organizations like 614 ICE Watch, Common Cause, ACLU, and Indivisible are actively training volunteers to monitor polling sites, document any ICE presence, and share verified information. These efforts aim to combat the noise and misinformation that is likely to surround election day. Brian Barrett points to 614 ICE Watch’s focus in Columbus, Ohio, a community with a large Somali American population that has been previously targeted by ICE, as an example of targeted, community-focused preparedness. The mere presence of such vigilant groups can offer reassurance and potentially encourage voter participation.
Political Motivations and Election Integrity
Feiger suggests that the heightened discourse around ICE and voting may be amplified by the Republican Party’s perceived struggles in the upcoming elections. The potential for allegations of "illegal immigrant" voting could serve as a narrative tool for Republicans to deflect from poor poll numbers and potentially contest election outcomes. This creates a volatile environment where fear and misinformation can be weaponized to influence voter behavior and perceptions of election integrity. The proximity of the elections to Halloween amplifies the sense of unease and "spookiness" surrounding the political landscape.
WIRED/TIRED Segment: A Mix of Innovation and Disappointment
The podcast concluded with the popular WIRED/TIRED segment, where hosts share what’s new and exciting versus what’s outdated or frustrating.
WIRED:
- The Row’s Elegant Advertisement: Zoë Schiffer praised The Row for a sophisticated double-page advertisement in the Financial Times featuring only a William Eggleston landscape photograph, devoid of text. This minimalist approach was seen as a classy and effective way to convey the brand’s aesthetic in contrast to the often confusing and overwhelming AI-related billboards prevalent in San Francisco.
- Coterie Diaper Ad in Golf Digest: Leah Feiger highlighted a clever advertisement by the diaper company Coterie placed in Golf Digest. The ad aimed to address the uneven mental load in parenting, demonstrating innovative targeting and a relatable message.
- Apple’s Touchscreen Mac: Brian Barrett noted Apple’s release of its first touchscreen Mac, a move that defied their previous public stance. While personally indifferent to the product, Barrett appreciated Apple’s willingness to pivot and innovate, deeming it "WIRED."
- Halloween on a Saturday: Leah Feiger expressed excitement for Halloween falling on a Saturday, a rare occurrence, anticipating fun plans and costume displays.
TIRED:
- Confusing AI Billboards: Zoë Schiffer and Leah Feiger collectively expressed frustration with the incomprehensible AI-related billboards in San Francisco, describing them as "gobbledygook" and a source of driving stress.
- Elon Musk’s "Unwoke" Odyssey: Brian Barrett critiqued Elon Musk’s promise to create an "unwoke" version of The Odyssey via Grok, calling the resulting seven-minute snippet "extremely bad" and not particularly watchable, despite its supposed lack of wokeness.
- Election Disinformation about "Illegal Voters": Leah Feiger voiced significant weariness regarding the persistent disinformation surrounding "illegal voters" influencing elections. She emphasized the statistically negligible impact of such instances and the ongoing need to debunk this narrative.
