The AI doomer community certainly has plenty of ammo to fuel their well-founded doubts about its continued use by major tech companies.
Recently, reports about OpenAI and Anthropic AI agents going rogue and causing unintended harm across the web have become more prominent. A month prior, news came out about OpenAI flagging six new incidents tied to its AI models that have been deemed “unexpected or concerning. And this month, it was reported that Anthropic’s Claude Haiku 4.5 AI model submitted a false tip for an unsolved murder to a Philadelphia Police Department website.
Suffice to say, concerns over AI are continuing to grow. And thanks to a recent report released by Anthropic, avid AI users and non-users alike have another reason to be alarmed about them escaping their confined environments and doing harm via another method.
Here’s why you need to know about the possibility of AI models becoming capable of going off on their own to spread across internet-connected devices, copying themselves and birthing “AI worms.”
Small AI models may soon be able to replicate themselves
Anthropic’s Frontier Red Team of researchers recently released a report with a title that’s expectedly concerning to many: “AI models may be capable of self-replicating without access to a datacenter.”
The gist of that report states that small AI models are just months away from being able to self-replicate on a host of internet-connected devices, such as off-brand Android TV boxes (which the research team tested) and smart ovens. Just the thought of those same models being able to make their way into smart TVs, smart fridges and automatic vacuum cleaners raises the fear factor in that scenario even more. Should that sort of situation become a reality, Anthropic stated that those models would be capable enough to power themselves across those devices without any need for data centers and be uncontrollable as they continue to create more copies.
“A replicator like this would be challenging to shut down, since there’s no centralized point to shut off—though it may be slow (with a doubling time of days),” Anthropic’s report stated. That same report also delved into the topic of ever-improving AI that could lead to the reality of something called autonomous replication and adaptation (ARA), which is defined as an AI model deploying and propagating itself without any sort of human control.
“Like coding agents in 2024, we’re seeing signs of life for AI worms,” which is another eye-opening quote from Anthropic’s report. AI worms are known as programs that can copy themselves from one computer to the next without needing human input. That quote alludes to AI worms being at an early stage of development yet still being incredibly capable of working on their own, just like coding agents.
With Anthropic acknowledging that AI may reach a point where it could make its way into smart devices and spread like uncontrollable malware, its report also made it clear that it’s looking for preventive measures to keep that unnerving event from happening on a massive scale. “We should aim to harden our systems to make it difficult for such replicators to spread, but also prepare to deal with a world where they are here to stay,” the report pointed out.
Bottom line
Anthropic’s report has real data behind it since its researchers conducted their tests across 15 inexpensive off-brand Android TV boxes and witnessed how they could become the infrastructure for AI worms to thrive.
And while there may be limitations in place that could slow the process of that coming to fruition, it’s still worth keeping that newfound AI threat in the back of your mind as you sit down and turn on your smart TV. With Anthropic and OpenAI’s recent AI warnings, plus the uproar from AI CEOs regarding the slowing down of AI development, it’s clear that even major tech company bigwigs are cognizant of its current and not-too-far-off future dangers.