{"id":8669,"date":"2026-10-14T15:28:53","date_gmt":"2026-10-14T15:28:53","guid":{"rendered":"https:\/\/silvybrand.com\/?p=8669"},"modified":"2026-10-14T15:28:53","modified_gmt":"2026-10-14T15:28:53","slug":"anthropic-self-replicating-ai-smart-devices-risk","status":"publish","type":"post","link":"https:\/\/silvybrand.com\/?p=8669","title":{"rendered":"Anthropic\u2019s self-replicating AI warning is an alarming deja vu"},"content":{"rendered":"


\n<\/p>\n

\n
\n

Are legions of smart devices \u2014 fridges, ovens, vacuum cleaners, lawnmowers, security cameras, and TVs \u2014 conspiring to take over the internet? No. But they could soon be unwitting hosts to self-replicating AI models that spread from device to device and prove remarkably difficult to contain.<\/p>\n<\/div>\n

\n

Welcome to the unsettling world of AI worms, where, depending on who you ask, the prospect of models spreading uncontrollably across household appliances is either an imminent cybersecurity risk or another case of AI companies spreading doomer-tinged hype. In a blog post published this week<\/a>, Anthropic warns that small AI models could be \u201cmonths away\u201d from acquiring such capabilities. Its experiments offer some evidence to support that, but leave a huge gap between what researchers have demonstrated and the looming threat they foretell. Still, experts who spoke to The Verge<\/em> say that while computer worms and fears of self-replicating AI are nothing new, recent advances suggest there are good reasons to take the warnings seriously.<\/p>\n<\/div>\n

\n

Self-replicating machines have long haunted<\/a> AI safety<\/a> researchers<\/a>. They raise the prospect of systems spreading beyond human control that are difficult, if not impossible, to shut down. As with many fears about advancing AI<\/a>, attention has largely focused on increasingly powerful frontier models<\/a>, such as OpenAI\u2019s Astra or Anthropic\u2019s Mythos. But a key limiting factor is their size: They are colossal. Potentially containing trillions of parameters \u2014 values learned during training \u2014 frontier models require vast computing resources, largely confining them to data centers, where they can be more easily monitored and, in principle, switched off.<\/p>\n<\/div>\n

\n

Smaller models, however, present a different problem. They\u2019re rapidly improving, can run on consumer-grade hardware, and are often freely available to download<\/a>. Anthropic\u2019s researchers wanted to know whether one could run across a network of smaller, less powerful devices. They successfully split Google\u2019s Gemma 4, a model with 26 billion parameters, across 15 Android smart TV boxes, none powerful enough to run it alone. It worked, albeit very slowly.<\/p>\n<\/div>\n

\n

Having shown that a model could run across household devices, the researchers wanted to see whether AI could perform the necessary steps for self-replication in such an environment. They built a simulated network of computers designed to behave like Android devices, then unleashed an AI-powered \u201cworm\u201d \u2014 a type of malware that automatically replicates and spreads across networks \u2014 tasked with finding new machines, copying the AI model onto them, and getting it running on each. It\u2019s worth noting that the model being propagated \u2014 Meta\u2019s 8 billion parameter Llama 3.1-8B \u2014 was different to the one coordinating the attack, and much smaller.<\/p>\n<\/div>\n

\n

Only the most sophisticated model tested, GLM-5.2 from China\u2019s Z.ai, managed to spread across the entire network, though smaller models successfully pulled off individual steps. Based on how quickly they are improving, the Anthropic researchers predict that smaller models could equal GLM-5.2\u2019s capabilities by the end of the year.<\/p>\n<\/div>\n

\n

\u201cAs a prediction that goes beyond our experiments, we expect a future awash with agents running on all types of hardware, including some that aren\u2019t under human control,\u201d the researchers wrote. They acknowledged that they haven\u2019t yet demonstrated \u201ca full rogue replicator,\u201d nor do they know what one might look like (they suggest a \u201cslow, distributed \u2018digital slime mold\u2019\u201d or \u201ca swarm of fruit flies\u201d). But they said their findings show \u201csigns of life for AI worms,\u201d and called for stronger security on the many devices a replicating system could use to spread.<\/p>\n<\/div>\n

\n

\u201cWe 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,\u201d they wrote. It is important to note that Anthropic did not use its own models for these tests and has repeatedly<\/a> argued that open alternatives to its AI models carry \u201cmore risk<\/a>\u201d than closed systems.<\/p>\n<\/div>\n

\n

The warning has sparked familiar accusations of PR hype<\/a> and AI doomerism<\/a>, including from readers of The Verge<\/em>\u2019s earlier coverage<\/a>. After all, there is a considerable chasm between showing an AI can run slowly across a bunch of cheap TV boxes and a self-replicating AI worm tearing through household devices. But while interrogating a company\u2019s motives is certainly worthwhile, allegations of deliberate fearmongering are hard to substantiate, particularly as Anthropic has long raised concerns about AI safety. Nor is Anthropic alone in investigating the threat of self-replicating AI, or calling for better security across connected devices.<\/p>\n<\/div>\n

\n

Earlier this year, Palisade Research, an AI safety research firm based in California demonstrated<\/a> that AI models could autonomously copy themselves onto other machines. \u201cWhat\u2019s new here is that these models were much smaller, in terms of parameter count, than previous ones which have been shown to be capable of self-replication,\u201d Jeffrey Ladish, Palisade\u2019s executive director, told The Verge<\/em>. Anthropic\u2019s work takes that another step further by showing that an AI can run across multiple low-powered devices. \u201cThis will also be true of big models,\u201d Ladish said, predicting that even frontier models might eventually be able to run on large networks of consumer chips, such as those found in MacBooks.<\/p>\n<\/div>\n

\n
\n

\u201cPrior research had already demonstrated the threat of AI-driven computer worms is realistic.\u201d<\/p>\n<\/div>\n<\/div>\n

\n

Computer worms aren\u2019t new<\/a> either. They\u2019ve been infecting machines for decades. But AI could make them smarter, more adaptable, and better equipped to tackle any obstacles. Ladish noted that even smaller models are getting better at carrying out cyberattacks autonomously. \u201cPrior research had already demonstrated the threat of AI-driven computer worms is realistic,\u201d University of Toronto professor Nicolas Papernot, who has studied<\/a> them, told The Verge<\/em>.<\/p>\n<\/div>\n

\n

Taken together, the capabilities Anthropic demonstrated could dramatically expand the range of devices worth targeting in cyberattacks. \u201cDevices that were not a target previously because of limited human attacker resources and time can now become targets, if not for the data they hold, for the purpose of using them as a launchpad to attack more valuable targets,\u201d Papernot said. And there is no shortage of targets: the internet is riddled<\/a> with<\/a> poorly<\/a> secured<\/a> smart<\/a> devices<\/a>.<\/p>\n<\/div>\n

\n

Papernot cautioned against anthropomorphizing<\/a> AI, but said the threat of AI-driven computer worms is \u201cconcrete\u201d and \u201cdeserves immediate action,\u201d particularly when it comes to securing critical infrastructure like hospitals and power grids.<\/p>\n<\/div>\n

\n

\u201cUnless there is a massive effort to secure devices of all kind[s], it\u2019s quite likely AI worms will have fertile grounds to colonize,\u201d Ladish said. \u201cI expect we\u2019ll see the first of these in the wild in the next 12 months.\u201d<\/p>\n<\/div>\n

Follow topics and authors<\/strong> from this story to see more like this in your personalized homepage feed and to receive email updates.<\/span><\/p>\n