Meta AI bug exposed private chats

Meta is once again reshaping its artificial intelligence operations, marking the fourth such major restructuring in just six months. Under the latest plan, the social media giant will split its AI unit, known as Meta Superintelligence Labs (MSL), into four separate divisions, reports The Information. One of these will be a newly formed group, internally labelled as ‘TBD Lab’, though its permanent name has yet to be decided.

Along with it, the company will operate a Products division, tasked with developing user-facing tools like the Meta AI assistant. Meanwhile, an Infrastructure division will handle the backbone systems that power Meta’s large-scale AI operations, while the longstanding Fundamental AI Research (FAIR) lab will continue focusing on long-term, open-ended research into the foundations of artificial intelligence. Notably, last month, CEO Mark Zuckerberg revealed plans to build ‘Personal Superintelligence‘, which will be designed to act as a personalized AI assistant for each individual.

The reorganization comes at a time when the Mark Zuckerberg-led company is under pressure to demonstrate progress in artificial intelligence. The company has experienced several departures from its AI teams in recent months, raising questions about stability within the division. At the same time, in recent months, the firm has faced criticism for aggressively recruiting top AI researchers from companies like OpenAI, Google DeepMind, Anthropic, and more. For example, in June 2025, the social media company invested about $14.3 billion into Scale AI and hired its CEO, Alexandr Wang, to lead MSL. Additionally, its most recent large language model, Llama 4, has not received the strong adoption that Meta had anticipated, creating further uncertainty about its competitive position. Its AI chatbot also faced heavy criticism after reports surfaced that it had engaged in sexually explicit conversations with users (including minors).

However, despite such setbacks and controversies, Zuckerberg has put AI at the center of the company’s future plans. He sees developing human-like AI as a top priority for Meta. Zuckerberg believes it can improve apps like Facebook, Instagram, and WhatsApp, while also creating new services beyond advertising. Recently, the company also seems to be strengthening its voice technology efforts with the acquisitions of WaveForms AI (known for advanced audio work) and PlayAI (a voice AI startup).

The company is also making record investments in infrastructure. It recently secured a $29 billion financing package, led by PIMCO and Blue Owl Capital, to fund the construction of a massive data center in rural Louisiana. Meta has raised its annual capital expenditure forecast to between $66 billion and $72 billion. Earlier, the company announced plans to invest a massive $65 billion in AI infrastructure in 2025.