Major AI companies have united in support of open weight AI models to promote accessibility, competition, and safety through transparency, while Anthropic stands alone opposing this movement, citing concerns over irreversible risks and misuse. Critics accuse Anthropic of prioritizing its closed business interests, highlighting a broader industry divide between open innovation advocates and controlled, corporate-managed AI development.
A rare moment of unity has emerged in the artificial intelligence industry, with major companies like Nvidia, Microsoft, Meta, Google, OpenAI, Amazon, and others publicly supporting the release of open weight AI models. These models allow users to download, inspect, and adapt AI systems independently, fostering competition, reducing costs, and expanding access across various sectors. However, Anthropic, the company behind the Claude AI model, notably refused to join this coalition, sparking widespread criticism. Many accuse Anthropic of using AI safety concerns as a pretext to protect its closed business model and limit open competition, effectively placing the future of AI behind corporate gates.
Open weight models differ from closed models in that their learned parameters can be freely downloaded and used without permission from the original creators. This openness enables greater independence for users, such as hospitals and startups, who can operate AI systems without relying on a single provider. The industry letter supporting open weights argued that American AI leadership depends not on one company owning the smartest model but on AI being widely accessible across industries and institutions. It also challenged the notion that closed models are inherently safer, emphasizing that open models allow broader scrutiny and development of safeguards.
Anthropic’s refusal to endorse the open weights letter positioned it as the sole major AI lab opposing the industry consensus. Critics, including prominent venture capitalists and AI experts, suggest that Anthropic’s stance is motivated by financial interests, as open models threaten its ability to charge premium prices for closed access. This tension intensified with the release of Kimi K3, a powerful open weight model from the Chinese company Moonshot AI, which allegedly used outputs from Anthropic’s Claude model to improve itself. This incident highlighted the risks Anthropic associates with open weights, such as unauthorized copying and competition.
In response, Anthropic’s CEO Dario Amodei clarified that the company does not advocate for a blanket ban on open weights but supports targeted regulations. These include restricting powerful computing hardware exports to authoritarian regimes, cracking down on large-scale unauthorized model copying (distillation), and mandatory safety testing for highly capable models. Anthropic’s concerns focus on irreversible risks posed by open weights, such as the inability to recall dangerous models once released, and the potential for misuse in cyberattacks or biological threats. However, critics argue that these measures could stifle open innovation and disproportionately benefit closed labs like Anthropic.
The controversy reflects a fundamental divide in AI’s future: one path favors controlled, accountable labs managing AI access and safety, while the other champions open models that democratize AI but carry risks of misuse. Both sides claim to prioritize safety but also support business models aligned with their positions. Anthropic symbolizes the controlled frontier approach, emphasizing irreversible risks, while the open model movement warns against concentrated corporate power controlling intelligence. This ongoing debate will shape whether humanity ultimately owns AI technology or rents it from a few dominant companies.