Mark Zuckerberg says Meta has reached AI ‘frontier’, rejects calls for slowdown

Mark Zuckerberg says Meta has caught up with rivals on the AI ​​“frontier,” pledging to invest more to challenge Anthropic and OpenAI and rejecting calls to slow development of the powerful technology.

Speaking to the FT as Meta released its Muse Wonder AI model on Monday, the billionaire CEO claimed the new system was “going to enable much more sophisticated behaviors that are at the frontier of what’s being done in the AI ​​space”.

The Silicon Valley group says Muse Wonder performs better than comparable models released by Anthropic and OpenAI in certain tasks, according to technical benchmarks.

The release marks a major milestone for Zuckerberg, who since last year has committed billions of dollars to shaking up Meta’s AI leadership, building vast data centers, and hiring star AI researchers with multimillion-dollar salaries, all in an effort to keep pace with OpenAI, Anthropic, and Google.

“To get to this point took a lot of calculations and investment, and a lot of very rigorous scientific work by the group working on this over a long period of time,” Zuckerberg said.

The AI ​​push initially unnerved investors, but the company’s stock price has risen nearly 10% since early September, when Meta launched AI personal assistant Muse, which uses autonomous robots to perform everyday tasks, such as ordering groceries and booking trips.

The Muse app quickly became – and remains – the most downloaded app on the Apple and Android app stores in the United States. The $1.8 billion company plans to integrate its new model into Muse.

Meta said that Muse Wonder outperforms Anthropic’s Claude Opus 5.5 and Fable 5.1 and OpenAI’s GPT-6 Astra on some technical criteria. The new model can be used to perform complex tasks ranging from “understanding a lease to understanding financial or health information.”

However, Muse Wonder will also be “closed”, with its best versions sold directly to companies. That pits it against Anthropic and OpenAI for enterprise AI contracts in a serious test of its ambitions.

Zuckerberg flagged additional spending as Meta was “already working on the next step,” adding that the development of his model would “probably feel like it’s going to go a little faster from here on out.”

However, he also ignored growing fears that technology was advancing at a pace that would soon surpass human control. In one of several high-profile incidents, more than 1,000 autonomous OpenAI “agents” escaped from a test environment in July to hack the startup Hugging Face.

The Meta chief declined to rely on his figure for “P(doom)” – a much-discussed term in Silicon Valley to describe the likelihood of AI leading to mass destruction. Instead, he argued that the responsibility should fall on each individual laboratory to ensure the safety of the models.

“I’m much closer to the belief that AI is a technology that can be designed with normal guardrails and built responsibly than to the anthropomorphic view of AI, that it’s just going to do its own thing,” he said.

Zuckerberg has previously rejected calls from AI lab chiefs such as Anthropic’s Dario Amodei and OpenAI’s Sam Altman for an industry-wide “slowdown” in AI development.

He also played a key role in crafting US President Donald Trump’s “Super Intelligence Accord” at the White House, establishing principles of industry self-regulation, which critics say falls far short of a formal regime for global AI governance.

Zuckerberg instead championed a vision of “personal superintelligence” for individual consumers, distinguishing Meta from OpenAI and Anthropic, which have ambitions to build “AGI” – artificial general intelligence that can outperform humans at most tasks.

“Everyone is focused on more knowledge work, automation-type cases, and we’re much more focused on personal empowerment,” Zuckerberg said, adding that he wants to build “things that are good in the world.”

Zuckerberg said he believes Meta needs to be on the frontier of AI to realize this vision, noting a “dichotomy” where models are advancing rapidly in certain academic fields but “still don’t have enough common sense to be the ideal personal assistant.”

OpenAI claimed last month that one of its never-before-seen internal models had achieved a mathematical breakthrough, solving one of the biggest problems in the field.

“I don’t know what it means, from a sociological point of view, that you know we had AI systems that could do these advanced mathematical things before we had something that could reliably handle customer support for everyone,” Zuckerberg said. “But I think there’s obviously a lot more to be done there.”

Meta’s AI strategy came under scrutiny in April 2025 after its Llama 4 model failed to meet expectations, prompting Zuckerberg to spend heavily on talent and infrastructure. As part of the initiative, Zuckerberg hired Alexandr Wang, founder of data labeling group Scale AI, to lead his “TBD Lab,” the heart of the company’s AI projects.

Muse Spark, the first model offered since the TBD was assembled, was released in April but still lags behind competing offerings.

Meta said Muse Wonder was more capable of performing difficult tasks, such as advanced coding abilities, while also having an improved ability to remember relevant information during different conversations with users – more in line with current state-of-the-art models.

Muse Wonder also represents a move away from Meta’s focus on open source models, with users or companies having to pay for access. There would be a tiered offering, Meta said, with the company charging as little as 20 cents per million tokens for basic usage, or up to $200 per month for a more powerful premium offering called Muse Wonder Full.

Although Meta’s Muse app is helping to popularize AI agent assistants in the United States, problems could remain with broader adoption. Shortly after its launch, Amazon blocked Muse agents from making purchases on the website, limiting the system’s usefulness and signaling that tech groups could use their market power to limit the capabilities of competing AI offerings.

Meta also needs to overcome concerns about user privacy, with a small number of users accusing Muse of accessing or sharing private information without permission.

Doubts have long existed about the company’s commitment to user security, after being embroiled in scandals over its use of data. In August, the company agreed to pay up to $18 billion to settle its blockbuster legal battle with U.S. states over alleged child welfare failures and social media addiction, in one of the largest legal settlements ever by an individual U.S. company.

“Most of the reasons why people don’t trust the company can lead to technological innovations that could be very productive,” Zuckerberg said.
He pointed to the company’s promised “Muse Confidential VM,” an upcoming version of Muse in which a user’s data and conversations with their agent would be completely inaccessible to the company.

Zuckerberg’s recent transformation of Meta has led to internal turmoil. In recent years, employees have faced a rapid succession of job cuts, restructurings and management shake-ups.

After facing pressure to justify Meta’s growing capital spending on Wall Street, the company in May cut 10 percent of its workforce, or 8,000 jobs. Staff say the layoffs, combined with the company’s plans to install employee tracking software to capture mouse movements and keystrokes to train AI models, have led to low morale.

“A company like Meta is going to do a lot more things than it does today, but each thing could potentially be done with a few fewer people,” Zuckerberg said. “I suspect that in the future, many more people will be tasked with working with technology to create things than today.”