The Fight Over Open Source AI, Anthropic's $1.5B Payout, NYC Socialists: Evictions = Violence?

The ongoing debate about open source versus proprietary AI models is intensifying amid rapid advances in Chinese open source development and U.S. policy uncertainty. The conversation highlights regulatory capture attempts by leading American AI labs, their rhetorical inconsistencies regarding “distillation,” and the broad economic stakes of open source AI for U.S. competitiveness, enterprise costs, and the diffusion of innovation.

Threat From Chinese Open Source Models and Policy Responses

The release of Moonshot AI’s Kimi K3, a Chinese open source model, has pushed U.S. policy circles into active debate. Kimi K3 is described as on par with some cutting-edge American models like Opus 4.8 and GPT 5.6, with claims of a 50% cost advantage. The launch caused a “panic,” with commentators initially fearing that China had not only caught up but was able to offer advanced models far more cheaply. However, further cost analyses indicate that Kimi K3’s operational savings are moderate, and its top-tier performance is largely in specific tasks—such as coding for web development—not across the board.

The White House has become involved, reportedly considering restrictions or outright bans on Chinese open source AI models. Axios reported that White House officials are exploring this, while internal disagreements persist, with some preferring to incentivize U.S.-based open source development over imposing bans. No final decision has been made; the administration remains divided over how best to safeguard American interests while allowing open AI collaboration and competition.

Reflecting market uncertainty, Polymarket odds of a U.S. ban on open source models jumped from 22% to 45% in a matter of days following K3’s launch, underscoring the unpredictable regulatory environment.

Flawed Logic Of Banning Open Source to Address Distillation

Both Anthropic and OpenAI cite concerns about “distillation”—the process by which one model learns from the outputs of another—as a reason for seeking government intervention. However, they stop short of accusing Chinese developers of intellectual property theft, highlighting a rhetorical inconsistency. If protecting from distillation is the goal, a ban on American access to Chinese models is illogical; the policy should logically ban Chinese access to American models. Instead, the proposed restrictions would isolate the U.S. AI market, forcing American enterprises to buy expensive proprietary solutions, weakening domestic competitiveness as global enterprises use cheaper open source alternatives.

David Sacks and Chamath Palihapitiya argue that the better solution is to require stronger Know Your Customer (KYC) protocols and account management at the service level. Rather than sweeping restrictions, frontier labs can enforce stricter user verification, slowing model access for distillers and solving the problem without sacrificing ecosystem openness.

Hypocrisy of Regulatory Capture Attempts by Frontier Labs

Anthropic, which has grown from $10B to over $70B in annual revenue, exemplifies that major U.S. AI companies have thrived without government protection. Despite stellarly successful business models, Anthropic and OpenAI lobby for protection from open source and foreign rivals, creating a double standard: they claim rights to use all published content while simultaneously opposing Chinese use of their models’ outputs.

Meanwhile, American entrepreneurs are actively leveraging Chinese open source models for innovation and new applications, revealing the artificiality and self-interest behind the push for regulatory capture.

Economic and Market Implications of Open Source Proliferation

A fundamental shift is underway: ...

Recent legal developments and ongoing disputes highlight how artificial intelligence companies handle copyrighted and proprietary materials as training data—and the contradictions and emerging strategies shaping the industry.

Anthropic Settlement and Distillation as Standard Practice

Anthropic Settled $1.5 Billion Copyright Lawsuit Over Pirated Books; Proceeds Go To Authors and Lawyers

Anthropic recently agreed to a $1.5 billion settlement to resolve a major copyright lawsuit, the largest of its kind in the United States. The lawsuit centered on Anthropic’s use of 7 million pirated books sourced from websites like LibGen to train its Claude AI model. Authors will receive $3,000 per covered book, and lawyers will receive $101 million of the settlement. Of the 500,000 books included in the settlement, 91% of authors have already claimed their compensation. This marks the first major AI copyright settlement, with many more anticipated in the future. The courts have previously indicated that AI training on copyrighted books may be considered legal under fair use, but this legal question remains unsettled.

David Sacks explains that Anthropic’s main legal fault was using pirated books without even buying a single copy. Sacks points out that had Anthropic purchased just one copy of each book, they could have argued for fair use—an unresolved issue still being litigated for AI data with both Anthropic and OpenAI defending the practice under fair use doctrine. They maintain that purchasing one copy of proprietary materials is fair use for training, distinguishing it from outright theft, such as stealing proprietary model weights.

Distillation Is Common Across Industries Where Companies Study Competitors' Outputs to Inform Development, Not Regarded As Ip Theft or Illegal Behavior

Chamath Palihapitiya and David Friedberg frame “distillation”—the process of analyzing competitors’ outputs to inform one’s own developments—as an industry standard, not limited to AI. Benchmarking competitors' outputs occurs in technology (such as search engine ranking comparisons at Google) and manufacturing (such as car companies modeling after each other's products). Friedberg notes that using a competitor’s publicly available outputs to refine one’s own product is standard practice and not equivalent to stealing proprietary algorithms or source code.

David Sacks and Jason Calacanis note that only when copyright violations are blatant—such as wholesale use of pirated material—do legal judgments enter. Otherwise, studying public outputs, even at “industrial scale,” remains a gray area, although Anthropic did coin the term “industrial scale distillation attacks” to describe the practice.

Contradictions in Proprietary Model Companies' Legal Positions

Ny Times Sues Openai for Unauthorized Training on Copyrighted Content; Openai's Defense Parallels Chinese Companies' Use of Public Outputs for Derivative Models

Ongoing lawsuits, most notably The New York Times’ suit against OpenAI, reveal significant contradictions in leading AI firms’ legal positions. OpenAI is being sued for scraping and training on New York Times articles without permission, in violation of the publication’s terms of service. OpenAI’s defense closely mirrors arguments used by Chinese companies who have trained their models using the outputs of American systems—claiming that learning from publicly available output is legitimate.

Meanwhile, OpenAI and Anthropic argue in other venues that training on global content, including proprietary data, is legitimate under fair use, provided they do not directly copy model weights or steal code.

Frontier Labs Deemed Distillation Ip Theft For Their Models, While Asserting Rights to Train On Global Content Without Permission

Frontier AI labs in the United States and China find themselves in a paradoxical spot: accusing one another of IP theft when their outputs are used for training derivative models (distillation), while simultaneously defending their own right to train on the world’s content. David Sacks points out this “Spiderman meme” scenario, highlighting the mutual accusations and lack of clear legal distinction.

Anthropic Avoided the Term "Ip Theft" In Their Post on "Industrial-Scale Distillation Attacks," Suggesting It Wouldn't Withstand Legal Scrutiny

Sacks notes that despite Anthropic’s introduction of the term “industrial scale distillation attacks,” the company deliberately avoided framing this as “IP theft” in their communications. This is due to the risk such an admission would pose to their own fair use cases. The companies argue national security and ethical reasons against ...

Unprecedented Investment by Cloud and AI Companies

Leading tech companies are investing at scales never seen before, shifting away from traditional strategies like buybacks and dividends to prioritize infrastructure needed for the AI era. Google forecasts an astounding $195-205 billion in capital expenditures ([restricted term]) this year alone, representing nearly 20% of the entire U.S. military budget according to external estimates. Tesla is also demonstrating aggressive expansion, with [restricted term] up 140% year-over-year and projecting $25 billion in infrastructure investments. SpaceX's parent company is similarly boosting its infrastructure commitments.

These enormous investments have resulted in both Google and Tesla reporting negative free cash flow, a rarity especially for Google, which is free cash flow negative for the first time since going public. This shift takes investors by surprise, but industry analysts believe these outlays will generate outsized long-term rewards. The consensus is that investing now in core AI and cloud infrastructure lays the groundwork for exponential future growth, even if current cash flows are negative.

Google's Differentiated Advantages in AI Infrastructure Competition

Google’s sustained 32% average return on invested capital over 25 years is seen as a justification for investor confidence in the company’s ability to deploy capital effectively. This consistent discipline leads many to give Google the benefit of the doubt on its current [restricted term] surge.

A key growth area is Google Cloud Platform (GCP), which has achieved 82% yearly growth and is running at a $100 billion annual rate. Google’s advantages include being model-agnostic: enterprises can deploy any AI model or workflow on Google’s infrastructure without being locked in. This flexibility allows Google to serve a fragmented marketplace where both proprietary and open-source AI models are proliferating.

This fragmentation is viewed as a structural tailwind for Google. As Chamath Palihapitiya and Jason Calacanis discuss, the more AI models and applications that emerge, the greater the demand for Google's underlying infrastructure, particularly at the silicon and cloud provider layers. Even if Google’s application-layer and consumer-facing models underperform, the company stands to monetize the world’s best infrastructure for years to come by processing, storing, and serving AI workloads for others.

In addition to GCP, Google’s breadth of assets—consumer platforms, YouTube, stakes in companies like SpaceX and Anthropic, Waymo, and a host of “other bets”—positions it to capitalize on multiple angles of the AI revolution, compounding its returns further.

Market Reaction and Valuation Implications

Despite the strategic rationale, markets have reacted negatively to this [restricted term] surge, largely because many investors are unaccustomed to infrastructure spending cycles o ...

Philosophical Foundation of Property Rights in American Democracy

David Friedberg references John Quincy Adams, emphasizing that the foundation of American democracy lies in private property rights. In his 1787 work, Adams warns that disregarding the sanctity of property leads to anarchy and tyranny, sentiments echoed in his 1791 essays. Friedberg articulates that early Americans sought refuge from monarchies where property could be seized arbitrarily by the ruling class. The promise of the United States was that individuals could own property free from such government overreach. He warns that when governments label property rights as illegitimate or begin small interventions—such as imposing new restrictions on property owners—it sets a precedent that destabilizes the protection of these rights and leads society towards authoritarianism.

Friedberg continues that socialist movements, such as those aligned with the Democratic Socialists of America (DSA), aim to systematically limit landlord authority through new regulations. These measures include restricting credit and background checks, changing income verification standards, and imposing stricter eviction procedures. He argues such policies make profitable and responsible property management significantly more difficult and that each incremental policy erodes the basic liberties enshrined by America's founding philosophies.

Negative Effects of Rent Control and Eviction Restrictions in NYC

David Sacks and Jason Calacanis highlight the practical impacts of these policies in New York City. Recent laws restrict landlords’ ability to conduct credit checks or verify sufficient income; landlords can require one but not both, which, they argue, limits their discretion to select reliable tenants. Other regulations bar landlords from evicting non-paying or problematic tenants, leaving little recourse for property owners.

The result, they state, is that many landlords opt to keep units vacant rather than risk renting to tenants they cannot vet or remove for nonpayment or delinquency. Calacanis and Sacks reference reports of 50,000 so-called "ghost apartments" in New York City—units kept intentionally empty by landlords. They argue this practice exacerbates housing shortages and drives overall rents higher.

Moreover, unmanaged problematic tenants in rent-controlled buildings cause significant harm to neighbors—especially elderly and working-class residents—through noise, safety issues, and general disorder. Maintenance deteriorates as landlords, deprived of rent income, lack funds for building upkeep, accelerating the decline of living conditions for all tenants.

Economic Solutions Ignored by Progressive Policy Advocates

Instead of addressing the supply side of the housing equation, progressives focus on regulatory restrictions for property owners. Calacanis and Chamath Palihapitiya maintain that basic economics dictate that increasing housing supply lowers rents, regardless of whether the units are luxury condominiums, multifamily buildings, or single-family homes. They cite evidence from cities like Austin, where permitting reform increased the number of new units and directly decreased rents. Similarly, Tokyo and Buenos Aires have succeeded through increased supply and deregulation.

In contrast, cities like New York and states like California are constrained by permitting barriers that stifle new construction, keeping ...