After AI slop flooded the internet with low-quality, machine-generated content, a new AI-related concern, ‘AI slime mold,' is taking over the web. AI slime molds are small rogue AI models identified by Anthropic's researchers that could spread across internet-connected devices and replicate themselves.
These rogue AI models could potentially find their way into household appliances such as smart ovens, refrigerators, robotic vacuum cleaners and lawn mowers as early as the end of 2026, according to a report by Business Insider.
In a report published on Monday, Anthropic's AI capabilities research team warned that many internet-connected devices have “minimal security measures in place”. This could allow small AI models to use their computing power to replicate and spread across devices.
The researchers described the potential scenario as a slow, distributed “digital slime mold” living on relatively insecure computing hardware.
The team tested the idea using several small AI models, including China's GLM series from Z.ai and Alibaba's Qwen, as well as Google's Gemma model from Google DeepMind. In one experiment, researchers ran inference using the Gemma model across 15 inexpensive, off-brand Android TV boxes, each costing $55, and found that the setup worked.
ALSO READ: Millions Of Records, Dozens Of Companies: Inside Japan's Escalating Cyber Crisis
What Is An AI Worm?
Worms have existed in cybersecurity for decades. These malicious programmes can copy themselves and spread across connected systems without requiring someone to manually install them on every device.
Anthropic's researchers warned that AI worms could take this threat further by replicating themselves on other devices while carrying their AI models along. Rather than relying entirely on powerful data centres, such systems could potentially use the computing capacity of everyday smart appliances.
The researchers cautioned that shutting down these distributed systems could prove difficult, particularly if they spread across devices that are poorly secured or not directly controlled by their owners.
"As a prediction that goes beyond our experiments, we expect a future awash with agents running on all types of hardware, including some that aren't under human control," the Business Insider report quoted the researchers.
They urged developers to strengthen cybersecurity measures while preparing for a future in which autonomous AI agents may become increasingly difficult to contain.
The warning comes amid growing concerns about AI safety. Last month, current and former employees raised alarms about the possibility of an AI catastrophe within the next decade, intensifying scrutiny of frontier AI companies, including Anthropic and OpenAI.
ALSO READ: Microsoft Bets On In-House Maia Chips for AI From 2027 To Loosen Nvidia's Grip
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.