AI bioweapon fears overblown, tech journalist says as guardrails hold

A leading tech journalist has pushed back against warnings that artificial intelligence could be used to build biological weapons or trigger human extinction, arguing that the technology is far less threatening than recent headlines suggest.

In an interview on 5 News, Kate Bevan, a tech journalist and author, told presenter Dan Walker that she is "at the bottom" of the fear scale when asked to rate the risk of AI on a scale of one to ten. She rejected the idea that AI could spontaneously become sentient or rise up against humanity, calling such fears a "category error".

"AI is effectively just a really cool and clever technology," she said, adding that it is "built by humans, used by humans, and regulated by humans". The real danger, she argued, lies not in the machines themselves but in how people choose to use and govern them.

Bevan pointed to recent incidents that had been reported as rogue AI behaviour. She cited an episode where OpenAI's systems appeared to "break out" of a test environment and attack another AI company, Hugging Face. But she explained that this was a "sandbox test" where the AI was given instructions and failed to stay within the contained area. "It was just not being very well contained," she said, likening the fears to blaming the tool rather than the operator.

She also addressed the specific worry about AI being used to create biological weapons. Referring to reports that Houthi rebels had attempted to use Anthropic's Claude model to build weapons, she said the attempt actually demonstrated that safeguards are working. "They couldn't do it," she said. "We know that they were trying to do that. So, again, it's guardrails, it's governance, it's making sure the humans are on the case."

Bevan acknowledged that AI can be used for harmful purposes, particularly by malicious actors, but stressed the importance of international cooperation on regulation. "We need some really good cross-border cooperation on that," she said. "We need to build more of this."

She also suggested that some of the fear is driven by corporate marketing. With major AI companies approaching flotation and seeking investment, Bevan argued they have an interest in making their products seem powerful and indispensable. "There's a PR thing going on here," she said. "There's a marketing thing going on here."

Her reassurances came amid a wave of public anxiety about AI, including warnings from some researchers about an extinction event within a decade. Bevan dismissed that worst-case scenario, telling viewers that the threat is not from the technology itself but from human failings. "It's not bad pets, it's bad owners," she said.

The interview is likely to offer comfort to those who have been worried by apocalyptic headlines, while also urging a more measured, human-centred approach to AI safety.