The rapid race to build more powerful artificial intelligence systems is moving so fast that the world may soon struggle to keep control says Anthropic co-founder Jack Clark, who has called for stronger safeguards to slow AI development when necessary.
Speaking during an interview with BBC Newsnight, Clark warned that the technology is approaching a critical threshold where AI systems could increasingly develop and improve themselves with minimal human involvement, raising profound questions about regulation, accountability and societal control.
“You want the option to be able to take your foot off the gas and put your foot on the brake,” Clark said. “Right now, it’s like the AI industry has a gas pedal, but it doesn’t have a brake pedal.”
His remarks follows growing global concerns over the breakneck pace of AI innovation, as companies compete to build increasingly sophisticated models capable of performing tasks once considered uniquely human.
Clark argued that governments and policymakers must play a greater role in ensuring that humans remain firmly in charge of technologies that are becoming more capable and influential across nearly every aspect of society.
“The world needs to do some thinking and we need to eventually develop some new regulations that allow us to be confident in these systems,” he said.
One of the most striking revelations from Clark’s interview involved Anthropic’s flagship AI assistant, Claude. According to him, about 80 per cent of the code behind the system is now generated by AI itself rather than written directly by human programmers.
He suggested that within the next two years, AI-generated code could account for virtually all of the software development process behind such systems, a shift he believes would have far-reaching implications for the technology sector and society at large.
Recall that the prospect of AI writing and refining its own code has become a focal point in debates about the future of artificial intelligence. While supporters argue it could dramatically accelerate innovation and productivity, critics warn it could also make AI systems more difficult to understand, regulate and control.
