Anthropic Just Became AI's Public Enemy #1

An unprecedented alliance just formed in the cutthroat world of artificial intelligence, uniting fierce rivals against a common foe. Over 235 companies and organizations, including industry titans like Google, Meta, OpenAI, and NVIDIA, signed the "Open Weights and American Artificial Intelligence Leadership" letter. This rare moment of consensus defended open-weight AI, a foundational shift for the entire sector.

But one major player conspicuously refused to join: Anthropic. Their absence from this historic coalition immediately became the biggest story, positioning the company behind Claude as the sole leading AI lab standing against an almost unanimous industry front. Anthropic's silence turned it into AI's public enemy #1, facing accusations of protecting its closed business model.

At the heart of this schism lies a critical distinction between how we access intelligence. When users interact with closed models like Claude or ChatGPT, they merely rent that intelligence. The creator company owns the servers, dictates terms, and can revoke access at will, trapping users in its ecosystem.

Open-weight models offer true ownership. Developers download their learned numerical parameters, running the models on their own infrastructure, adapting them, and inspecting their inner workings without requiring permission. This empowers businesses with independence, fostering innovation beyond corporate gates:

• Hospitals maintain sensitive data on-premises.
• Universities conduct research without per-request fees.
• Startups build products free from permanent dependency on a single provider.

The accusation against Anthropic is blunt: their high-minded appeals to AI safety serve as a convenient smokescreen. Critics contend these concerns are a calculated maneuver to hobble cheaper, more accessible open-weight alternatives threatening their lucrative, closed business model. This isn't about protecting humanity; it's about protecting market share.

Venture capitalist David Sacks didn't mince words. He warned Anthropic aims to "kneecap" open source, orchestrating regulatory capture to cement its competitive advantage. The industry, he argued, must monitor the company "like a hawk," lest it corner the market on the guise of responsible AI development.

China’s Kimi K3 model vividly illustrates this nightmare scenario for closed labs. Developed by Moonshot Artificial Intelligence, Kimi K3’s powerful open-weight release shocked Silicon Valley. This near-frontier model, available for free, obliterates the scarcity premium that allows companies like Anthropic to charge exorbitant prices for proprietary access.

A competitor releasing a top-tier model that developers can download and adapt destroys the very foundation of a closed system's pricing power. This economic reality fuels the suspicion that Anthropic's safety posturing is less about genuine risk mitigation and more about erecting an impenetrable moat around its commercial interests.

Anthropic CEO Dario Amodei insists his company never advocated a blanket ban on open-weight models. He champions open models lacking "dangerous capabilities" as a vital public good, benefiting developers and researchers alike. Amodei's core concern, however, centers on the catastrophic, irreversible risks from future, extremely powerful systems, explaining Anthropic's refusal to sign the industry's unifying Open Weights and American AI Leadership letter. He believes that document makes "dangerous assumptions" about the implications of such potent models becoming widely downloadable.

Amodei articulates two chilling nightmare scenarios. First, an authoritarian government could achieve military superiority, mass surveillance, or permanent repression through AI more powerful than American systems. Second, a sufficiently potent model might enable widespread cyberattacks or the creation of biological weapons.

To mitigate these risks, Anthropic proposes "targeted regulations." These include:

• Restricting advanced chip sales to China
• Cracking down on industrial-scale model copying
• Implementing mandatory safety testing for frontier AI systems

Critics, however, contend these measures would effectively choke open-source competition. They argue the proposals impose prohibitive costs and introduce crippling legal uncertainty, benefiting only well-resourced closed-model developers like Anthropic.

This is the central struggle for AI's future: will intelligence be controlled by a few accountable corporations, or distributed for anyone to own, adapt, and inspect? This profound question defines the coming era, pitting centralized corporate control against the radical transparency and decentralized empowerment of open weights.

Anthropic’s current position invites pointed charges of hypocrisy from across the industry. Critics recall perceived past instances where the company argued for greater access and less governmental restriction when its own closed models encountered US government scrutiny. This stance now appears directly contradictory to its opposition to broad open-source release of powerful models, fueling the industry's backlash against its perceived self-interest.

Dario Amodei’s defense, however, centers on a potent, if chilling, vision for future AI. He contends that a dangerously capable open model is inherently irreversible; its weights, once released, cannot be recalled, making containment impossible. Such a system, Amodei fears, risks "irreversible catastrophic risks" from the most powerful future systems, potentially unleashed globally without a safety net.

The open movement, conversely, posits an equally compelling counter-argument: a handful of corporations should not be the sole gatekeepers of intelligence. They champion the right for anyone to own, adapt, and inspect these foundational technologies. This distribution fosters robust competition, ensures auditability by a wider community, and prevents a future where humanity merely rents its most critical capabilities, undermining true American AI leadership. The ultimate prize is control over intelligence itself.

Why is the AI industry suddenly against Anthropic?

Nearly every major AI company, including Google, Meta, and OpenAI, signed a public letter supporting open-weight AI models. Anthropic, a key player, conspicuously refused to sign, leading to accusations that it's using safety concerns to protect its closed, for-profit business model from open-source competition.

What are open-weight AI models?

Open-weight models allow developers to download the model's numerical parameters. This enables them to run, inspect, and adapt the AI on their own systems, free from the control and pricing of the original creator, unlike closed models like ChatGPT or Claude which are accessed as a service.

What is Anthropic's official argument against releasing powerful open-weight models?

Anthropic CEO Dario Amodei argues that releasing extremely powerful models is irreversible and could lead to catastrophic misuse, such as development of bioweapons or use by authoritarian regimes for military superiority. They propose targeted regulations, not a blanket ban.

What is industrial-scale distillation in AI?

Distillation is a technique where one AI model learns from the outputs of another. Anthropic is concerned that rival companies, particularly from China, could use this technique on a massive scale to copy the capabilities of their expensive, proprietary models like Claude at a fraction of the cost.