Meta CEO Mark Zuckerberg Just Published a 6,500 Word Essay Taking Direct Aim at OpenAI and Anthropic

The artificial intelligence war shows no signs of slowing.

Earlier on Aug. 10, Meta Platforms (NASDAQ:META) CEO Mark Zuckerberg released a 6,500-word essay that appears to take direct aim at the large language model (LLM) leaders OpenAI and Anthropic.

Tension has been building between Zuckerberg and OpenAI and Anthropic. In a recent New York Times article, Zuckerberg criticized the two companies for centralizing power in the space in a way that harms innovation and poses risks to the sector.

In Zuckerberg's latest essay, while the longtime CEO doesn’t explicitly name OpenAI or Anthropic, he appears to take clear aim. Here is what Zuckerberg said.

The ongoing debate about open and closed models

Zuckerberg and a slate of other prominent tech and artificial intelligence CEOs have come out in favor of open-source AI models.

In fact, a group of hyperscalers and other prominent AI companies, including Nvidia, recently issued a joint letter urging lawmakers not to move too quickly in regulating open-source AI models, which can be accessed and modified by anyone.

OpenAI and Anthropic run proprietary, closed-loop models, which have thus far generated the best performance and are believed by some to be safer from a regulatory perspective. However, they are also more expensive than open-source models.

Chinese companies have recently released LLMs with better performance and that are reportedly 60% to 90% cheaper than proprietary models.

Zuckerberg’s essay comes as Meta launches a group of open-source models called Muse Glimmer, specifically aimed at laptop use. Meta is also planning to open the weights on its latest AI model, Spark 1.2.

The stock has struggled this year as the company ramps up capital expenditures, but has left investors questioning its AI strategy.

“I do not believe restricting access to foreign open source models is an effective solution,” Zuckerberg wrote in his essay. “Our goal should be for American open source models to be the best globally. This requires removing the hurdles that make it harder for American open source models to compete.”

Zuckerberg also argued that open-source systems have proven safer from a cybersecurity standpoint because there are more people who can “… identify vulnerabilities, harden the systems, and easily upgrade to the latest most secure versions.”

At one point in his essay, Zuckerberg appeared to hone in on Anthropic CEO Dario Amodei, who has previously expressed extreme caution about AI, arguing that there needs to be strict regulation and taxes, or it could disrupt society as we know it.

“I do not understand why anyone who believes that AI will eliminate most jobs and much of humanity’s relevance would rush to build that future,” Zuckerberg wrote. “The notion that AI is so dangerous that the only safe path is an extreme concentration of power seems inherently problematic.”

Zuckerberg likely has ulterior motives, but it’s a real debate

While Zuckerberg may very well believe what he is saying, it is also good for Meta’s business if regulation on open-source models is less stringent, since Meta is planning to launch a batch of them.

But make no mistake: the open vs. closed debate is becoming a bigger part of the AI story, especially as enterprise-level AI token usage has become more expensive.

Furthermore, if open-source models are not overly restrictive and prove cheaper, they could erode the moats of OpenAI and Anthropic.

The question lawmakers will face is how risky these open models are from a cybersecurity and regulatory perspective. But how this debate shakes out is likely to have real implications for all of the major hyperscalers and AI companies, as well as their stocks.

Bram Berkowitz has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Meta Platforms and Nvidia. The Motley Fool has a disclosure policy.