Poaching Tactics Falling Short? Meta Poised to Embrace External AI Models

August 30 – By Zhu Ling, Sci-Tech Innovation Board Daily

Mark Zuckerberg has come to a clear realization: building world-class AI applications takes more than offering sky-high salaries to lure top talent. The real winning move lies in embracing cutting-edge AI technology—no matter where it comes from.

Earlier this month, reports surfaced suggesting that Zuckerberg had “conceded” in the race against ChatGPT, with Meta stepping back from directly competing in OpenAI’s core territory.

Now, fresh developments have added a new twist—Meta is reportedly exploring collaborations with none other than its long-time rivals, OpenAI and Google, to supercharge the AI capabilities across its product ecosystem.

Opening the Door to External Partnerships

Within Meta’s Superintelligence Lab, executives have discussed integrating Google’s Gemini or OpenAI’s models into Meta AI, the company’s chatbot platform, to answer user questions and support other AI-driven features across Meta’s social media apps. If these talks come to fruition, Meta would be embedding competitor technology directly into its flagship products.

A Meta spokesperson confirmed the company’s broad-based approach to AI leadership—combining in-house development, partnerships with external firms, and the adoption of open-source technologies.

In fact, Meta has already started incorporating outside models into its internal tools. Its coding assistant now offers both Anthropic and OpenAI models as selectable options for employees. Another internal productivity assistant has also been integrated with OpenAI’s technology.

Meanwhile, Meta’s Chief AI Officer, Alexander Wang, recently announced a licensing agreement with Midjourney, granting Meta access to its advanced AI-powered image and video generation technology.

Zuckerberg’s Strategic Dilemma

This pivot toward external collaboration underscores a deeper challenge for Zuckerberg. While Meta commands enormous influence in the social media world, its footing in the rapidly evolving AI arena remains far less secure—marked by urgency and, at times, reactive decision-making.

Earlier this year, development of Meta’s flagship AI model, Behemoth, hit major roadblocks, pushing back its launch date repeatedly. Zuckerberg was reportedly dissatisfied with the Llama 4 team’s progress, believing they failed to deliver the breakthroughs needed for Behemoth.

In response, he overhauled the AI leadership, bringing in former Scale AI CEO Alexander Wang and ex-GitHub CEO Nat Friedman to co-lead the Superintelligence Lab—backed by multi-billion-dollar investments.

Meta also aggressively recruited high-profile AI talent from Apple, Anthropic, xAI, Google, and OpenAI, seeking to close the gap with industry leaders.

The company is now developing its next-generation model, Llama 5, with the aim of going head-to-head with the likes of OpenAI’s GPT series and Google’s Gemini. However, such ambitious R&D takes time.

The Immediate Challenge

Meta AI faces a critical hurdle: despite a massive user base of one billion monthly active users, only about 10% interact with the platform daily. Until Llama 5 is ready, Meta needs interim solutions to boost product competitiveness—and tapping into external AI models may be the most pragmatic path forward.