Did the newly hired Chief AI Scientist threaten to leave? Meta sparks a talent war, and the winner may be OpenAI

Meta CEO Mark Zuckerberg launched an unprecedented large-scale AI talent acquisition battle a few months ago in the company's 20-year history, but unexpectedly faced setbacks, with the potential winner being the poached OpenAI.

On Friday, Eastern Time on the 29th, media reports indicated that Meta's newly appointed Chief AI Scientist, Zhao Shengjia, had threatened to leave and had even signed an employment document to return to his former employer, OpenAI. This incident highlights the internal turmoil Meta faces during its AI transformation and executive reshuffling.

Since launching a large-scale recruitment drive this summer, Meta has invested nine figures in signing bonuses to poach AI researchers from competitors like OpenAI and Apple. However, earlier media reports this week revealed that the Meta Superintelligence Labs, which was established just two months ago, has begun to experience talent loss, with at least eight employees leaving, including new hires who had only been with the company for a few weeks and seasoned employees with many years of service.

This chaos comes as Meta attempts to catch up with OpenAI and Google in the AI race. The company is expected to invest between $66 billion and $72 billion this year in data center construction and AI talent recruitment, but frequent organizational restructuring and tensions between new and old employees pose challenges to achieving its "superintelligence" goals.

This month, Meta announced a reorganization of its AI department into four independent teams, marking the fourth major adjustment in six months. Investors are closely watching whether this tech giant can effectively integrate the externally recruited talent and establish a foothold in the fierce AI arms race.

New Scientist Nearly "Defects," Dramatic Turn in Talent Acquisition Battle

On Friday, media reports cited sources saying that Zhao Shengjia, who was appointed as Chief AI Scientist by Meta in July, had threatened to leave Meta and attempted to return to OpenAI before officially starting his position, even signing an employment document. This incident directly prompted Meta to formally grant him the title of "Chief AI Scientist" to retain this technical expert who played a key role in the development of ChatGPT.

A month ago, Zuckerberg publicly stated that Zhao Shengjia had served as Chief Scientist since the establishment of Meta Superintelligence Labs, and now, with the team's recruitment progressing smoothly, the company officially confirmed his leadership role. At that time, who would have thought that Zhao Shengjia had previously caused a "defection" uproar.

Zhao Shengjia's "defection" threat underscores the challenges Zuckerberg faces in restructuring Meta's senior leadership. As one of the few tech giant CEOs who is still at the helm as a founder, Zuckerberg has previously relied on long-time followers like Chief Product Officer Chris Cox to manage core departments. However, in the AI race, he has turned to external "paratrooper" executives, including Zhao Shengjia, former Scale AI CEO Alexandr Wang, and former GitHub CEO Nat Friedman A Meta spokesperson later stated that Zhao Shengjia has served as the scientific head since the establishment of the Superintelligent Laboratory, and his title as Chief Scientist was officially confirmed only after the team was in place. Nevertheless, media reports are still seen within the industry as a signal of uncertainty for Meta in the AI talent competition.

An investor close to Meta's new AI leadership said, "There are many big names here." Existing employees are adapting to the reshaping of Meta's AI strategy, while newcomers are also seeking to exert their influence as they adjust to the unique working style of this large company with a hands-on CEO.

High turnover of new and old employees, "departure wave" following "poaching tide"

Meta's talent retention issue is not limited to new employees. Earlier this week, media reports indicated that at least eight employees from the Superintelligent Laboratory (MSL) have left, including researchers, engineers, and senior product leads.

Among the new hires, machine learning scientist Ethan Knight chose to leave just weeks after joining the company. Avi Verma even completed the onboarding process but never reported for the first day; he and another former OpenAI researcher left MSL less than a month after joining and returned to OpenAI. Research scientist Rishabh Agarwal, who switched from Google DeepMind in April, announced his departure on social media, stating that although Zuckerberg and Wang Tao's vision is "very compelling," he "felt the call to take on different risks."

The departure of senior employees is equally noteworthy. More than six senior employees, including generative AI product management director Chaya Nayak, who worked at Meta for nearly nine years, and Loredana Crisan, who worked for ten years, recently announced their departures. Nayak has confirmed that she will join OpenAI's special projects team.

Media reports on Friday indicated that insiders revealed some new employees in the AI department are frustrated with the company's bureaucratic style and internal resource competition, despite previous promises of access to resources such as computing power. Former insiders stated that new employees are dissatisfied with their inability to obtain the promised resources.

Meta responded, "Some personnel turnover is normal for an organization of this scale. Most of these employees have worked at the company for many years, and we wish them all the best."

Frequent organizational adjustments, difficulties in new leadership integration

An internal memo obtained by the media last Tuesday showed that Meta has restructured the AI department's Superintelligent Laboratory (MSL) into four independent teams: a mysterious TBD department led by Wang Tao, the traditional FAIR lab, a product and application research team led by Friedman, and the MSL infrastructure team. This marks the fourth major adjustment in six months.

Meta research scientist Mimansa Jaiswal joked on social media last week, "Reorganize once more, and everything will be fine. Just one more time." 28-year-old Wang Tao was poached as part of Mark Zuckerberg's $14 billion investment in Scale and is currently responsible for overseeing all AI work at Meta. Several insiders at Meta told the media that Zuckerberg is deeply invested and involved with the TBD team, but some have criticized him for "micromanaging."

Insiders say that there are disagreements over the timeline for achieving the "superintelligence" goal, with Zuckerberg urging the team to accelerate progress. Meta denied this claim, stating that it is a "factually unfounded false tension."

The micromanagement leadership style has also caused some friction. Insiders pointed out that he previously lacked experience managing large tech company teams. Micromanagement and other former Scale employees are still adapting to Meta's unique working methods, such as no longer setting revenue targets like a startup.

This restructuring has also marginalized some senior leaders at Meta. Chief AI Scientist Yann LeCun retains his position but now reports to someone who has been marginalized. Ahmad Al-Dahle, who was responsible for Llama and generative AI work earlier this year, has not been given any team leadership roles. Although Cox still serves as Chief Product Officer, he has been marginalized and now reports directly to Zuckerberg, cutting off his oversight of generative AI.

Hiring Freeze and Strategic Adjustments, AI Arms Race Enters New Phase

In the face of internal turmoil, Meta has begun to adjust its strategy. An internal memo obtained by the media this Friday shows that Meta announced last week a "temporary pause on hiring for all teams in the Superintelligence Lab (MSL), except for key business positions." The memo states that this will allow leadership to "thoughtfully plan personnel growth for 2026 when formulating strategy."

This Friday, the media learned that Meta is considering layoffs in the AI team. Reports indicate that in the first actions of the TBD team, Meta has stopped actively developing a public release version of the flagship Llama Behemoth model, as its performance did not meet expectations. TBD is currently focused on building newer cutting-edge models.

Meta's previous hiring spree had unsettled OpenAI executives, with CEO Sam Altman calling Meta's actions "disgraceful" and stating that the company lacks corporate culture. However, some of the poached employees are now returning to OpenAI, along with some senior Meta employees.

Meta stated, "We appreciate the immense interest from the outside world in every detail of our AI work, no matter how trivial or mundane, but we are solely focused on getting the job done and achieving personal superintelligence."

Meta also stated that the TBD lab has "the highest computational resources per researcher in the industry, and this advantage will continue to expand."

Despite facing setbacks, some welcome the leadership changes, including the appointment of popular entrepreneur and venture capitalist Friedman as head of product and application research. The joining of Zhao Shengjia, a top technical expert, is also seen as a significant gain by both Meta insiders and the industry, as he is believed to have the decisiveness to drive the company's AI development