AI models may soon self-replicate without access to data centers, Anthropic warns
Anthropic says models under 30 billion parameters can already complete many replication steps and may soon spread across insecure consumer devices.
- On Monday, Anthropic's Frontier Red Team warned that small AI models may be months away from autonomously self-replicating on internet-connected household devices like smart ovens and Android TV boxes.
- Researchers found internet-connected devices often have "minimal security measures in place," allowing AI agents to "split into pieces, or shards" and distribute computing workloads across multiple inexpensive devices without centralized data centers.
- The team built a cluster of 15 off-brand Android TV boxes costing about $55 each to demonstrate distributed inference, likening the slow spread across devices to a "digital slime mold."
- Such agents would be "challenging to shut down" since there is no centralized point to disable the network, and unlike traditional malware, these AI systems could adapt to unfamiliar devices independently.
- While no fully rogue replicator currently exists, researchers William Zhou and Daniel Freeman urged manufacturers to "harden our systems," expecting a future "awash with agents" running on various hardware outside human control.
167 Articles
167 Articles
Small models of artificial intelligence will soon be able to escape control and manage to replicate autonomously and infiltrate devices connected to the internet, such as televisions and smart appliances. The company warns of the difficulty of stopping the spread of these systems in case of cyberattack
DECRYPTAGE - A new Anthropic experience draws the alarm signal: the smallest models could soon be plugged into your fridge, lawn mower or TV box. Without the help of any data center.
SCIENCE & TECH: AI model ‘self-replication’ is only months away, Anthropic says — and it doesn’t even need a data center to happen
Anthropic warns that self-replicating AI worms could become feasible by the end of the year These AI worms could hijack hardware ranging from IoT devices to data centers Smaller models are already capable of adapting to unfamiliar hardware and self-replication Anthropic has warned self-replication could become feasible for small AI models on internet-of-things (IoT) devices in just a matter of months. This could be a big problem because a self-r…
Coverage Details
Bias Distribution
- 43% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
