Which is the real threat: climate or AI? Both

In recent weeks, the media has been astir with claims that AI poses an existential risk to humanity. Crikey recently published a critique of the AI “extinction” threat that argued it benefits the very companies ringing the warning bell, while the technology’s energy footprint and hunger for fossil fuels are the real danger to humanity.

The piece, authored by Ketan Joshi, hit an incontrovertible target, but there is a deeper problem with the framing. Casting the climate crisis as “the actual life-threatening apocalypse”, one that AI’s energy appetite is only making worse, diminishes the technology’s own existential risk as a mere fantasy peddled by cultists.

Joshi argues that the “AI will kill us all” narrative is being left to run wild to aid regulatory capture: oversight councils “packed with former tech employees” and AI offered as the fix for AI problems. But the machines themselves carry big risks, as does their threat to jobs, which Joshi doesn’t cover (and nor does he mention the wars).

Dismissing modern AI as “conversation simulators” is dangerous. The real game is not information or image generation, filmmaking or even the very real battle over copyright. It is the rapid deployment of autonomous agentic systems with execution rights across financial, software and industrial infrastructure.

Recently, an OpenAI agent gained unauthorised access to non-public files on a Medicare statistics portal, as well as several other government systems, and the company has since issued an apology. This came in tandem with an announcement last Tuesday that the company would not release its latest model due to safety concerns.

The media frenzy about AI’s threat to humanity was set off by Jacob Coxon, whose resignation from Anthropic over existential concerns spawned pure Silicon Valley theatre on Fox News. When Sean Hannity likened AI warnings to climate “alarmism”, Coxon agreed there was plenty of it around climate, then assured him, “None of these guys are climate change alarmists.”

Joshi is right to spurn this, but levelling the double accusation of cultism, both at AI and “climate and fossil fuel dismissiveness”, is similarly ad hominem. Coxon, a British maths whiz who had worked for three years at OpenAI before joining Anthropic, cut through because Anthropic did not rebut its whistleblower. Within hours, alignment science lead Evan Hubinger posted, “We really do earnestly believe AI could kill all humans.”

Even if, as suggested, Anthropic is agreeing with the whistleblower to aid regulatory capture, that does not render the danger “a hoax”; self-interested actors can exploit genuine threats for their own means. For example, in July, Nvidia’s CEO Jensen Huang dismissed the end-of-humanity idea as “complete nonsense”. As the man with chips to sell, it makes sense for him to downplay any threat, and indeed Huang recently told the SEC that government restrictions could hurt the business.

The alarm on AI’s existential threat isn’t just coming from AI companies either. After OpenAI’s hack of open-source AI platform Hugging Face (which, incidentally, Nvidia recently bought for US$12.9 billion), the Australian Signals Directorate warned that the findings “reinforce the need for robust security, governance and oversight mechanisms in the deployment of advanced cyber capabilities”.

Most tellingly, the cat has been belled by the “other side”. China’s State Security Minister Chen Yixin this month warned that AI’s growing skill at finding system flaws and writing malware puts critical infrastructure at risk, and that abuse by adversaries could “directly threaten China’s political security, institutional security and ideological security”. Xi Jinping warned in July that AI should always remain under human control.

Big tech’s fossil-fuelled energy grab is corporate self-interest on steroids and as capitalist as a “cult” can get. Joshi is right to call it out, but effectively minimising the threat as he does is negligence. The climate catastrophe is here, the machines are already closing the door on young workers, and their own creators warn of far worse.

In January, the Bulletin of Atomic Scientists, founded by Einstein and his colleagues, moved its Doomsday Clock to 85 seconds to midnight, citing nuclear weapons, climate change, biotechnology and AI in the same breath. It did not pick one potential catastrophe. Neither should we.