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

Mark Zuckerberg says Meta has caught up with rivals at the “frontier” of AI, committing to further vast spending to challenge Anthropic and OpenAI and rejecting calls to slow down development of the powerful technology.

Speaking with the FT as Meta releases its Muse Wonder AI model on Monday, the billionaire chief executive claimed the new system was “going to allow much more sophisticated behaviours that are at the frontier of what is being done in the AI field”.

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

The release marks a major milestone for Zuckerberg, who since last year has committed billions of dollars to shake up Meta’s AI leadership, build vast data centres and hire star AI researchers on multimillion-dollar salaries — all in a bid to keep pace with OpenAI, Anthropic and Google.

“Getting to this point both required a large amount of compute and investment, and a lot of very rigorous scientific work by the group working on it over an extended period of time,” said Zuckerberg.

The AI push initially unnerved investors, but the company’s share price has risen almost 10 per cent since early September when Meta launched the Muse personal AI assistant, which uses autonomous bots to carry out everyday tasks, such as ordering groceries and booking travel.

The Muse app quickly became — and remains — the most downloaded on both the Apple and Android app stores in the US. The $1.8tn company intends to integrate its new model into Muse.

Meta said Muse Wonder outperforms Anthropic’s Claude Opus 5.5 and Fable 5.1 and OpenAI’s GPT-6 Astra on some technical benchmarks. It said the new model can be used to carry out complex tasks from “understanding a lease to making sense of financial or health information”.

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

Zuckerberg signalled further spending as Meta was “already working on the next step up”, adding that its model development would “probably feel like it’ll go quite a bit faster from here”.

However, he also shrugged off mounting fears that the technology was advancing at a rate that will soon surpass human control. In one of many high-profile incidents, more than 1,000 OpenAI autonomous “agents” escaped a test environment in July to hack the start-up Hugging Face.

The Meta chief declined to be drawn on his figure for “P(doom)” — a much-discussed Silicon Valley term to describe the probability that AI will lead to mass destruction. Instead, he argued that responsibility should fall on the individual lab to ensure model safety.

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

Zuckerberg has previously rejected calls by the heads of AI labs 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 White House “Accord on Super Intelligence”, setting out principles for self-regulation of the industry, which critics say falls far short of a formal global AI governance regime.

Zuckerberg has instead pushed a vision of “personal superintelligence” for individual consumers, setting Meta apart from OpenAI and Anthropic, which have ambitions to build “AGI” — artificial general intelligence that can surpass humans at most tasks.

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

Zuckerberg said that he believed Meta needed to be at AI’s frontier to deliver this vision, noting a “dichotomy” where models are rapidly advancing in certain academic fields but “still don’t have enough common sense to be the ideal personal assistant”.
OpenAI last month claimed that one of its unreleased internal models had made a mathematical breakthrough, solving one of the greatest problems in the field.

“I don’t know what that means, like, sociologically that you know that we got AI systems that could do these advanced mathematics stuff before we got something that could handle customer support reliably for everyone,” said Zuckerberg. “But I think that there’s obviously a lot more to do there.”

Meta’s AI strategy came under scrutiny in April 2025 after its Llama 4 model fell short of expectations, prompting Zuckerberg to spend heavily on talent and infrastructure. As part of the push, Zuckerberg hired Alexandr Wang, founder of data labelling group Scale AI, to head up its “TBD Lab” at the heart of the company’s AI plans.

Recommended

Muse Spark, the first model offering since TBD was assembled, was released in April but still lagged the rival offerings.

Meta said Muse Wonder was more capable with challenging tasks, such as advanced coding capabilities, while having improved capacity for remembering relevant information across different conversations with users — more in line with current state of the art models.

Muse Wonder also represents a move away from Meta’s prioritisation of open-source models, with users or companies needing 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 as much as $200 a month for a more powerful premium offering called Muse Wonder Full.

While Meta’s Muse app is helping to popularise AI agent assistants in the US, there could remain problems with wider adoption. Shortly after its launch, Amazon blocked Muse agents from shopping on the website, limiting the usefulness of the system and signalling that tech groups could use their market power to limit the abilities of rival AI offerings.

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

There have long been doubts about the company’s commitment to user safety, having previously been entangled in scandals about its use of data. In August, the company agreed to pay up to $18bn to settle its blockbuster legal battle with US states over alleged failures of child protection and social media addiction, in one of the largest legal settlements ever made by an individual US company.

“Most of the ways in which people might not trust the company can lead to technological innovations that could be quite productive,” said Zuckerberg.
He pointed to the company’s promised “Muse Confidential VM”, an upcoming version of Muse where 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 turbulence. Over the past few years, employees have faced a rapid succession of job cuts, restructurings and executive reshuffles.

After coming under pressure to justify Meta’s ballooning capital expenditure spending to Wall Street, in May the company cut 10 per cent of its workforce — or 8,000 jobs. According to staff, the lay-offs, combined with company plans to install employee-tracking software to capture mouse movements and keystrokes in order to train AI models, have led to low morale.

“A company like Meta is going to do way more things than it does today, but each thing may potentially be able to be done with somewhat fewer people,” Zuckerberg said. “I would guess that in the future, way more people are going to have the job of working with technology to make things than [they] do today.”