Meta is cutting around 600 jobs from its newly formed superintelligence division as the company reorganizes its sprawling artificial intelligence operations, signaling a shift toward smaller, faster teams in what CEO Mark Zuckerberg has called Meta’s “year of efficiency.”

In a memo to staff on Wednesday, Meta’s chief AI officer, Alexandr Wang, said the restructuring was meant to make the team “more agile and talent-dense” as the company pursues its ambitious goal of building “personal superintelligence” — a term Zuckerberg uses to describe AI systems that could eventually surpass human capabilities.

“By reducing the size of our team, fewer conversations will be required to make a decision, and each person will be more load-bearing and have more scope and impact,” Wang wrote.

The layoffs mark one of the most significant reorganizations within Meta’s AI division since the company established the Meta Superintelligence Lab (MSL) in June. The division was designed to spearhead research into next-generation AI models, consolidating engineers and scientists from multiple Meta units — including Facebook AI Research (FAIR), infrastructure, and product teams — under one structure.

Meta has described MSL as a cornerstone of its long-term AI strategy, with Zuckerberg pledging billions of dollars in compute and talent to compete with OpenAI, Anthropic, and Google DeepMind. The company has been aggressively recruiting since mid-2024, offering multimillion-dollar pay packages to lure researchers from rival firms.

Over the summer, Meta poached more than 50 scientists and engineers from competitors, including OpenAI and Google, though OpenAI CEO Sam Altman downplayed the departures. “None of our best people took the offers,” Altman said at the time.

Despite the hiring spree, the rapid growth of MSL has created internal strain. Current and former employees have described overlapping mandates and shifting priorities that led to confusion and tension across Meta’s broader AI organization. Several senior researchers reportedly left within months of joining.

The latest cuts appear aimed at resolving that internal friction by consolidating roles and tightening coordination. Wang’s memo said the reorganization would allow the division to “move faster” and empower each remaining team member to have greater influence.

“These are talented people who have worked extremely hard and contributed to our AI effort,” Wang wrote. “We are supporting the majority of those impacted in finding new roles at the company.”

He added that Meta had formed a “tiger team of recruiters” to help displaced employees quickly transition into new positions within the company.

Meta said the restructuring does not meaningfully reduce its overall headcount, as many of those affected are expected to be rehired in other AI or engineering roles.

The company’s move fits into a broader philosophy articulated by Zuckerberg since early 2023, when he began describing Meta’s “year of efficiency” as a way to make the company “leaner is better.” That initiative led to tens of thousands of job cuts across divisions, aimed at flattening management layers and improving decision speed.

In the case of the superintelligence lab, the goal appears to be organizational streamlining rather than cost-cutting.

“Our goal is to enable MSL to move faster,” Wang wrote, emphasizing that Meta’s investment in compute and AI infrastructure would continue. “We remain excited about the models we are training, our ambitious compute plans, and the products we are building, and I’m confident in our path to superintelligence.”

The superintelligence project has become one of Meta’s most expensive bets, with the company investing hundreds of millions of dollars to scale its compute clusters and train large AI models. Meta’s AI ambitions span multiple fronts — from the consumer-facing “Meta AI” assistant integrated into its apps to foundational models powering future augmented and virtual reality experiences.

Zuckerberg has framed AI as central to Meta’s long-term future, positioning the company as a challenger to OpenAI and Google in both research and product deployment.

“We want to build general intelligence openly and responsibly,” he said earlier this year. “Our goal is to make it available to everyone as widely as possible.”

Wang’s memo reflects that same ethos of aggressive efficiency paired with scale. His remarks echo a recurring theme across Meta’s AI units — that smaller, specialized teams are better suited to innovate in a field moving at unprecedented speed.

The restructuring also comes at a time when competition among leading AI companies has intensified. OpenAI, Anthropic, Google, and Meta are each training new frontier models that require massive amounts of compute power, driving a surge in demand for Nvidia chips and data center capacity.

Within this race, Meta has sought to differentiate itself by emphasizing open research and democratized access to its AI tools, including the Llama family of open-source models. The company recently launched Llama 3, which has been widely adopted by developers and integrated into several enterprise systems.

Still, internal tensions have emerged as Meta expands its ambitions beyond open-source models to the more exclusive superintelligence project, where research secrecy and proprietary models are expected to play a larger role. Some researchers reportedly expressed unease over the shift toward confidentiality and the heavy emphasis on rapid scaling.

Wang’s decision to trim the MSL workforce underscores the company’s effort to stabilize the division’s structure and clarify its direction after a summer of intense hiring and shifting goals. Analysts see the move as a sign that Meta is entering a more disciplined phase in its AI race — focusing on execution rather than expansion.

While the exact financial impact of the restructuring remains unclear, Meta insists that the investment pipeline for its AI research remains intact.

“This by no means signals any decrease in investment,” Wang reiterated. “We will continue to hire industry-leading AI-native talent.”

General Motors has announced plans to introduce a conversational AI assistant powered by Google’s Gemini across its cars, trucks, and SUVs starting next year, marking a significant leap in the automaker’s broader effort to fuse artificial intelligence with its connected vehicle ecosystem.

The announcement was made on Wednesday at the GM Forward event in New York City, where the company unveiled several upcoming technologies. The Gemini integration is expected to be the first among them to reach consumers, ahead of a planned automated driving feature and a revamped vehicle computing platform slated for release in 2028.

GM’s integration of Gemini builds on an already deep relationship with Google, whose “Google built-in” operating system currently powers infotainment systems in Buick, Chevrolet, Cadillac, and GMC vehicles. Those systems already allow drivers to access Google Assistant, Maps, and other Android apps directly from their dashboard screens.

In 2023, GM expanded that collaboration by using Google Cloud’s Dialogflow chatbot to manage non-emergency OnStar services such as navigation assistance and general inquiries. The forthcoming Gemini-powered assistant will take that concept further by leveraging generative AI to hold more natural, context-aware conversations with drivers.

“One of the challenges with current voice assistants is that they’re often frustrating — they rely on code words and struggle with accents or phrasing,” said Dave Richardson, GM’s senior vice president of software and services, in an interview with TechCrunch. “What’s great about large language models is they’re flexible in how you speak to them. They understand context from previous interactions, so you get a better, more natural experience.”

The Gemini assistant will be capable of managing a wide range of tasks — from drafting and sending messages and planning routes with multiple stops to accessing information from the web, such as answering a driver’s query about a historical landmark.

It will also integrate more deeply with OnStar, GM’s in-car concierge system, allowing it to provide maintenance alerts, route suggestions, and even control features like heating or cooling before the driver enters the car.

In essence, GM envisions the system as “a mix between a health wearable and an AI companion, but for your car,” Richardson said.

The assistant will be distributed over the air via the Google Play Store, and will be available to OnStar-equipped GM vehicles dating back to model year 2015 — significantly broadening access to existing customers.

Privacy and Data Concerns After Past Controversy

The announcement comes amid growing scrutiny of how automakers handle driver data. GM faced backlash earlier this year after reports revealed that the company had shared customer driving and geolocation data with insurance firms, prompting privacy concerns.

Addressing those concerns directly, Richardson emphasized that the new AI system will prioritize user consent and data protection.

“Everything that we’re going to do is going to be driven by customer consent, so you can always opt in or opt out,” he said. “Our viewpoint is that data and privacy have to be built into everything that we do.”

He also noted that GM has hired a new data governance team, including Christina Montgomery, who previously served as IBM’s chief privacy and trust officer, to establish stronger oversight mechanisms and compliance frameworks.

Richardson added that any driver data collected through Gemini “goes directly toward improving the product and won’t be sold for revenue.”

Toward a Custom-Built AI

While GM’s collaboration with Google is central to the Gemini rollout, Richardson said the company intends to test multiple AI models, including those from OpenAI, Anthropic, and others, as it moves toward developing a custom-built, domain-specific AI system trained on vehicle data.

“The idea is to take a base language model and train it on the vehicle’s specifications — distill that down and have it running locally on the vehicle,” Richardson said.

That approach would give GM more control over performance and data privacy, while enabling the AI to understand specific vehicle functions such as one-pedal driving or energy optimization in EVs.

GM’s move places it squarely in the middle of an emerging AI arms race among automakers. Competitors have already begun incorporating generative AI into their vehicles: Mercedes-Benz has integrated ChatGPT, Stellantis is working with French AI firm Mistral, and Tesla has rolled out Grok, the conversational assistant developed by Elon Musk’s xAI.

For GM, however, the Gemini rollout could serve as a bridge between existing OnStar services and a future where AI-driven personalization becomes a key selling point for vehicles. The assistant will enable the company to build a continuous feedback loop — learning from driver behavior while providing proactive assistance — potentially transforming how customers interact with their cars.

Analysts say that if executed effectively, the Gemini assistant could also enhance GM’s competitiveness in the electric and autonomous vehicle markets, where software-defined vehicles are increasingly defining brand differentiation.

With its rollout planned for 2026, GM’s Gemini-powered assistant is expected to mark a pivotal moment for the company’s digital transformation.