In a stark intervention that has sent ripples through the tech community, the co-founder of Anthropic has declared that artificial intelligence must never be allowed to develop without continuous human supervision. Speaking at a closed-door summit on AI safety, the executive warned that the rapid acceleration of machine learning capabilities could lead to a scenario where algorithms operate beyond our comprehension and control, a path he described as “catastrophic for civilisation.”
Anthropic, a company built on the principle of ethical AI, has long advocated for rigorous safety measures. But this latest statement carries particular weight because it comes from a figure who helped create some of the most advanced neural networks in existence. He argued that the current trajectory of AI development, fuelled by commercial pressures and a lack of global regulation, risks creating systems that could autonomously make decisions affecting everything from financial markets to military defence.
“We are building black boxes that even their creators cannot fully explain,” he said. “If we do not embed human oversight into the very architecture of these systems, we will lose the ability to steer them. The idea that we can ‘fix it later’ is a dangerous illusion.”
His comments echo a growing unease among Silicon Valley’s inner circle, where figures like Eliezer Yudkowsky and Nick Bostrom have long warned about the existential risks of AGI. But what makes this warning different is its timing. Just last week, OpenAI demonstrated a model that can write code, design websites, and pass university-level exams. Competitors are racing to match or surpass these feats, often prioritising speed over safety.
The Anthropic co-founder did not mince words. He called for a moratorium on the development of certain high-risk AI applications until international agreements are in place. He also urged governments to invest in interpretability research, which aims to make AI decision-making transparent to humans. “Without interpretability,” he said, “we are flying blind.”
Critics argue that such measures could stifle innovation and cede ground to authoritarian states with fewer ethical constraints. But the Anthropic executive dismissed this as a false choice. “The real competition is not between nations, but between wisdom and recklessness. If we allow AI to evolve unchecked, we won’t have a civilisation left to compete for.”
The speech comes amid a broader push for AI regulation in Europe and the United States. Just this month, the EU advanced its AI Act, which categorises applications by risk level. But enforcement remains patchy, and many companies continue to operate in a legal grey zone.
For the average person, these debates might seem abstract. But the implications are deeply personal. With AI already curating our news, managing our benefits, and diagnosing our illnesses, the question of who — or what — is in control is becoming urgent. The Anthropic co-founder’s warning is a reminder that the future is not something that happens to us. It is something we build, with every line of code and every policy decision we make.
As he concluded, his voice carried a mix of hope and dread: “The next few years will define the long arc of our species. We must choose to keep the needle of agency firmly in human hands.”
