Meta Considers Shifting AI Strategy from Open Source to Closed Models Amid Monetization Pressures

Meta, known for its strong commitment to open-source AI, is reportedly considering a significant shift in its strategy. According to The New York Times, top members of Meta’s new Superintelligence Lab are discussing the possibility of moving away from the company's powerful open-source AI model, Behemoth, and focusing on developing closed-source models instead. This pivot is driven by several factors, including underwhelming internal performance of Behemoth and the pressure to monetize AI technologies beyond advertising. Reports indicate that Meta completed training on Behemoth but delayed its release due to poor performance. The launch of the Superintelligence Lab further halted testing on the model. However, any final decision on this shift would require approval from Meta CEO Mark Zuckerberg. A company spokesperson told TechCrunch that Meta’s stance on open-source AI remains unchanged, noting that they will continue to release leading open-source models while training a mix of open and closed models. Zuckerberg has previously expressed ambivalence about fully committing to open-source AI. In a podcast last summer, he stated that while Meta is generally pro-open source, they would not release everything they develop if it posed responsibility concerns or if there was a significant qualitative change in the model's capabilities. Meta’s emphasis on open-source AI has been a key differentiator from competitors like OpenAI, which partnered with Microsoft and became more closed. The Llama family of models, in particular, has been touted as a more transparent alternative. However, despite its robust AI research capabilities, Meta has lagged behind companies like OpenAI, Anthropic, Google DeepMind, and xAI in commercializing its AI technologies. The shift to closed models could give Meta greater control and additional revenue streams, especially as the company invests heavily in talent, data centers, and AGI development. Top researchers are being poached with substantial signing bonuses and nine-figure salaries, highlighting the intense competition and financial stakes in the AI race. Industry insiders see this potential change as a strategic move rather than an ideological one. If Meta indeed prioritizes closed models, it could have wide-ranging impacts on the AI landscape. The momentum behind open-source AI, driven in large part by Meta’s Llama models, might slow, potentially giving a boost to larger, closed ecosystem companies. Smaller startups that rely on open-source foundation models for fine-tuning, safety, and alignment could face significant challenges. Globally, Meta’s retreat from open-source AI could cede ground to China, which has actively promoted open-source AI initiatives like DeepSeek and Moonshot AI to enhance domestic capabilities and expand its international influence. The ripple effects would extend to both the tech industry and geopolitical dynamics. In conclusion, while Meta maintains that its open-source commitment remains intact, the discussions around Behemoth and the focus on closed models indicate a growing recognition of the commercial and competitive pressures in the AI field. This potential shift underscores the evolving nature of AI development and highlights the balance between innovation, control, and profitability that companies must navigate. Rebecca Bellan, a senior reporter at TechCrunch, covers a wide range of tech topics including Tesla, AI, and Big Tech regulatory issues. Her previous work at Forbes.com and contributions to various publications reflect her expertise in technology and business. Bellan's insights provide valuable context to the ongoing discussion about Meta’s AI strategy and the broader implications for the tech industry.