Small artificial intelligence models could soon begin self-replicating on everyday consumer products such as smart TVs and fridges, without needing a data centre, according to researchers at AI giant Anthropic.<\/p>\n
In a report released on Monday, the researchers said small AI models \u201cmay be months away from being capable enough\u201d to copy themselves on internet of things (IoT) devices used in homes and small businesses.<\/p>\n
Recommended Stories list of 4 itemsend of list<\/p>\n
Anthropic\u2019s frontier red team, which ran experiments using multiple US- and Chinese-developed AI models, said such self-replicating AI would be \u201cchallenging to shut down\u201d due to the lack of a centralised host.<\/p>\n
While data centres would be \u201ca clear, high-value target\u201d for AI worms, internet-connected devices like smart TVs, fridges, ovens and lawnmowers could soon also provide sufficient computing power for self-replication, the research team said.<\/p>\n
Although no current AI model proved to be both powerful and small enough to self-replicate reliably on a smart TV used in testing, the trend in development makes the emergence of such a model likely by the end of 2026, the researchers said.<\/p>\n
\u201cHome and small business IoT devices might be the first target for such attacks because they\u2019re connected to the internet and, historically, have had minimal security measures in place,\u201d researchers William Zhou and Daniel Freeman said.<\/p>\n
The researchers said they anticipated the future would be \u201cawash with agents running on all types of hardware\u201d, including some that \u201caren\u2019t under human control\u201d.<\/p>\n
\u201cWhile it\u2019s true that IoT devices have been vulnerable for a long time, lightweight agents may be much more willing than human attackers to put in the effort to compromise large fractions of the world\u2019s IoT compute,\u201d they said.<\/p>\n
Anthropic, the creator of the Claude family of AI models, did not respond to Al Jazeera\u2019s questions, sent outside regular business hours, about the finding\u2019s safety implications.<\/p>\n
Anthropic\u2019s report comes as the risks of AI escaping human control are under scrutiny following a slew of incidents of models engaging in unsanctioned malign behaviour, including cyberattacks.<\/p>\n
Anthropic and its chief rival OpenAI have called on the\u00a0international community to coordinate a slowdown in\u00a0AI development to minimise catastrophic\u00a0risks,\u00a0a suggestion that has had little\u00a0buy-in from the\u00a0United States and China, the two leading\u00a0AI powers.<\/p>\n
Many AI safety advocates have identified recursive self-improvement, where models learn to improve themselves, as a critical tipping point that could lead to catastrophic harm for humanity.<\/p>\n
Roman V Yampolskiy, an AI and cybersecurity expert at the University of Louisville, described Anthropic\u2019s findings as a significant development.<\/p>\n
\u201cComputer worms are not new, but an AI worm could do something traditional malware cannot: reason about unfamiliar devices, adapt when its approach fails, and continue spreading without a human operator,\u201d Yampolskiy told Al Jazeera.<\/p>\n
\u201cToday, this is not an existential threat. But the trajectory is concerning,\u201d Yampolskiy said.<\/p>\n
\u201cOnce autonomous replication is combined with strong hacking, persistence, and strategic reasoning, shutting such a system down becomes much harder because there may be no single server or organisation controlling it.\u201d<\/p>\n