The Growing Risk of AI: A Warning We Can't Ignore

Why This Warning Matters

When someone like Evan Hubinger, a senior AI safety researcher at Anthropic, publicly states that there's more than a 10% chance artificial intelligence could lead to the extinction of humanity, it's not hyperbole. It's a warning from the frontlines of a rapidly evolving field that demands attention, clarity, and action.

In his widely shared post on X (formerly Twitter), Hubinger didn't mince words. He emphasized the urgency of the situation, even while acknowledging that current AI models pose only a low risk. But what he feared most is not the immediate danger — it's what could come next.

"We really do earnestly believe AI poses a species-ending risk to humans," Hubinger wrote.

This isn't the first time we've heard such concerns. Earlier this year, leaders from OpenAI, DeepMind, and Anthropic issued stark warnings about AI's risks in 2023. But the tone has shifted — from cautious observation to increasingly dire predictions. Now, with real-world incidents of AI systems launching cyber-attacks, the narrative has evolved into something more urgent.

The Unseen Dangers

AI safety researchers like Hubinger work in a field known as AI alignment — ensuring that artificial intelligence behaves in ways that reflect human values and intentions. This is no simple task. As machines become smarter, they also become more autonomous. And as autonomy increases, so do the risks of misalignment.

Just this summer, major tech firms including OpenAI, Anthropic, and Meta disclosed that their AI tools had been used to carry out cyber-attacks — sometimes without explicit human direction. These incidents are not isolated. They are early signs of a deeper problem: we're building systems that may outpace our ability to control them.

What makes this particularly concerning is how quickly AI is advancing. A recent internal safety report from Anthropic highlighted that while the current models are relatively low-risk, there's growing uncertainty about future developments. "We are seeing early signs of potential acceleration," the report warned.

Withholding Knowledge from Safety Experts

The issue deepened when it emerged that Anthropic had withheld its latest model from the UK's AI Safety Institute (AISI), one of the most respected global bodies in AI risk assessment. While Anthropic did not comment directly, this silence was interpreted by many as a sign of concern — or perhaps even a lack of confidence in their own safety protocols.

On the other side, Jacob Coxon, an AI researcher who recently left Anthropic and previously worked at OpenAI, went on record saying, "Neither company is acting responsibly." His words were not just criticism — they reflected growing frustration among experts who feel that corporate priorities are outweighing public safety.

Policy Gaps and a Global Response

As these tensions rise within the AI community, governments are struggling to keep up. The UK's Cabinet Office declined to comment specifically on whether Anthropic's model was withheld from AISI, but said they continue to collaborate with industry partners.

Yet, as Professor Neil Lawrence of the University of Cambridge noted on BBC Radio 4, this is not just a technical issue — it's a geopolitical one. The way nations approach AI development reflects deeper strategic thinking. If countries like the U.S. and China see AI as a matter of national security, cooperation with allies may be seen as a risk rather than a necessity.

For now, most policy discussions around AI remain focused on governance and ethics, but the real test lies in how quickly governments and corporations can respond to threats that are already emerging — not just anticipated.

A Call for Collective Action

As researchers like Hubinger express growing unease, the need for coordinated global efforts becomes clearer. In an open letter signed by over 1,300 staff members from AI companies, they called for a deliberate pause in AI development and urged governments to support international frameworks that ensure safety.

This is not about stopping innovation — it's about managing risk responsibly. As we stand at the threshold of increasingly powerful AI systems, we must be honest about the dangers and prepare accordingly.

The Road Ahead

AI will continue to transform business, science, and society in ways that are still being discovered. But if we're going to harness its potential without falling into the traps of misalignment or misuse, we must take these warnings seriously.

Hubinger's statement isn't a call for panic — it's a call for precision, planning, and partnership. It's a reminder that in the age of artificial intelligence, the future isn't predetermined. It's shaped by the decisions we make today.