Key Takeaways:
Meta CEO Mark Zuckerberg took direct aim at OpenAI and Anthropic, accusing the two AI leaders of pushing for centralized control over artificial intelligence development.
Meta Chief Executive Officer Mark Zuckerberg criticized OpenAI and Anthropic for their push to tightly control AI development, throwing his support behind open-source models as the industry's cost structure comes under scrutiny.
"The industry benefits from more openness, not less," Zuckerberg said in an interview published July 28. "Concentrating power over AI in a few hands creates risks that outweigh any safety benefits."
The comments come as OpenAI and Anthropic have separately lobbied Washington for stricter regulation of open-source AI models, according to people familiar with the matter. Both companies have argued that unrestricted access to powerful AI systems poses national security risks, a position Zuckerberg rejected as self-serving.
The clash represents a deepening divide in Silicon Valley over how AI should be governed — one with billions of dollars in implications. Meta has invested heavily in its open-source Llama model family, while OpenAI and Anthropic have raised tens of billions of dollars from investors including Microsoft Corp. and Amazon.com Inc. on the promise of proprietary, safety-controlled systems.
The Lobbying Battle
OpenAI and Anthropic have quietly pushed for rules that would require government licenses for the most capable AI models, a framework that would effectively lock out open-source competitors, according to regulatory filings reviewed by Bloomberg. The two companies have argued that open-source models could be used by bad actors to create bioweapons or launch cyberattacks — claims that Meta and other open-source advocates say are unsubstantiated.
Zuckerberg's public broadside marks an escalation of a conflict that has simmered for more than a year. Meta released its Llama 4 model in early 2026 with a permissive license that allows developers to modify and redistribute the software freely. By contrast, OpenAI's GPT-5 and Anthropic's Claude 4 remain closed systems accessible only through paid APIs or subscription products.
The Cost Factor
The debate over openness is colliding with a growing backlash against the cost of proprietary AI. A corporate fad known as "tokenmaxxing" — maximizing usage of AI tokens as a badge of productivity — is fading as companies confront rising bills without corresponding gains, according to a report from Moody's Ratings.
"Token costs have been doubling almost every other month," said Jue Wang, a management consultant at Bain & Co. who advises large enterprises on AI strategy. "At $200 per developer per month multiplied by 20,000 developers, that quickly becomes a line item no general manager planned for."
Palantir Technologies Inc. Chief Executive Officer Alex Karp told CNBC earlier this month that American businesses are "livid" about paying for tokens that create no value. Microsoft Corp. CEO Satya Nadella has warned that customers of proprietary models are paying twice — once for tokens and again by feeding their proprietary data to the AI provider.
Open-source models from Chinese startups such as Moonshot's Kimi and Zhipu's GLM have emerged as lower-cost alternatives, nearly matching the capabilities of top US models at a fraction of the price. That has intensified pressure on OpenAI and Anthropic to justify their pricing.
What's at Stake for Investors
Meta shares have gained 34% this year as investors reward the company's disciplined approach to AI spending. The company trades at 24 times forward earnings, a discount to Microsoft at 31 times and a premium to Alphabet at 20 times. If the open-source narrative gains further traction, it could pressure revenue growth at OpenAI and Anthropic, both of which remain privately held but have seen their valuation expectations rise on the strength of their proprietary models.
The regulatory outcome remains the biggest variable. If Washington adopts the licensing framework backed by OpenAI and Anthropic, open-source development could face significant hurdles. If it sides with Zuckerberg's vision of openness, the competitive advantage of proprietary models could erode.
This article is for informational purposes only and does not constitute investment advice.
