AI needs brake pedal: Anthropic’s Jack Clark warns AI could soon build itself, says ‘I am worried for my kids’

Jack Clark, co-founder of Anthropic, warned that AI is accelerating towards a threshold where it could develop more advanced versions of itself without human intervention, and called on governments to create regulatory mechanisms that could slow the technology’s progress before that time comes.

Statements made in separate meetings BBC news night And axiosis one of the most direct public warnings the AI ​​industry has made about the pace and direction of its own development.

Anthropic’s Jack Clark calls for ‘brake pedal’ on AI as industry moves forward

Jack Clark’s main argument is structural: the AI ​​industry currently has no mechanism to slow itself down.

“You want the option to take your foot off the gas and put your foot on the brake,” Clark said. BBC News Night. “Currently, the AI ​​industry has a gas pedal but no brake pedal.”

The Anthropic co-founder called on governments to take the lead in developing tools that can provide this control, arguing that public trust in AI depends on policy frameworks rather than the attitudes of individual company leaders.

“The world needs to do some thinking, and we need to develop some new regulations that will eventually allow us to trust these systems,” he said.

Drawing a historical analogy, Clark noted that the early oil industry was a template for how societies had previously turned to powerful and disruptive technologies.

“Society’s response has been to create a sensible policy and regulatory framework that gives people confidence in oil and the benefits it can bring to the world and means you don’t have to worry about the personalities of the people running companies,” Clark said. “It’s clear that’s where we end up here.”

Claude currently writes 80% of his own code; 100% will be possible within two years

The urgency behind Clark’s call is based on data emerging from Anthropic itself. The company has found that Frontier models already speed up coding, debugging, and research processes.

Anthropic’s Claude chatbot currently runs on code that 80% of the system wrote itself. Reaching 100 percent within two years is feasible and “will have huge implications,” Clark said.

Anthropic describes this trend as one that points to “recursive self-improvement,” a process in which AI systems autonomously design, build, test and improve their successors.

Clark detailed this phenomenon in an interview. axios: “In the near future, AI systems may become capable enough to autonomously design, build, and train more capable successors on their own. If this happens, each new version of Claude could be created by the previous version without human intervention.”

Clark warns that the progress of artificial intelligence will accelerate, not stagnate

Far from predicting a slowdown in the pace of AI development, Clark argued that the opposite is true.

“The big story here is that, contrary to some popular views, we are seeing indications that AI progress will accelerate rather than stay the same or decline in the coming years,” he said. axios.

This possibility is particularly important for scientific and medical research, but broader integration into existing industries requires careful preparation, he said.

“As organizations, and possibly societies, we need to find tools to verify and verify that the work done by these AI systems is accurate and aligned with human intentions that are compatible with a thriving society,” Clark said.

Anthropic has signaled that it plans to inform lawmakers about recursive self-improvement in the coming months, trying to engage lawmakers in the conversation before the phenomenon becomes widely reported.

“We’ve always found that the best thing to do is to socialize the concept and basically give people an idea of ​​what’s going to happen,” Clark said.

OpenAI has also published findings on recursive self-improvement, warning in a December 2025 blog post that it could pose a danger if AI research organizations do not openly share information on the subject.

Anthropic prepares for stock market debut as valuation approaches $1 trillion

The warnings come at a commercially important time for the company. Anthropic has grown significantly since it was founded five years ago by chief executive Dario Amodei, Clark and a small group of other executives. Private investors currently value the company at nearly $1 trillion, and it is set to list on the public exchange in a move that would rank among the most valuable stock market debuts in history.

Anthropic consistently positions itself as willing to openly discuss the risks of AI technology, even if it creates friction. The company has previously been involved in a public dispute with the US Department of Defense over concerns that its vehicles could be used in mass surveillance and autonomous warfare.

Clark was forthright about his personal motivation. “I worry about my children if we don’t have a serious conversation as a society about what the continued advances of artificial intelligence mean,” she told Newsnight. “There are potentially huge benefits. There are also risks.”

What does AI disruption mean for jobs and the economy?

Clark acknowledged that the economic consequences of the expansion of artificial intelligence are among the most pressing concerns he raises. Artificial intelligence agents, which are essentially autonomous robots that can perform routine tasks, are already cited by major technology companies as justification for large-scale workforce reductions.

On the question of human creativity, Clark measured: “There are open questions about whether AI systems can be truly creative… there is no real evidence of that yet. At Anthropic, we are now limited to the ability to generate good ideas rather than the ability to engineer those ideas into reality.”

Anthropic’s co-founder’s advice to young people worried about their prospects in an AI-driven economy was to go for the things machines can’t yet replicate. Clark suggested that young people should “develop a hobby” and pursue a liberal arts education.

“Those who benefit most from this are people who can think creatively and broadly, who read a lot and are interested,” he said. “Indulge your curiosity and learn how you can use this technology.”