\n<\/p>\n
A new warning about artificial intelligence from Anthropic caused a stir in cybersecurity circles on Tuesday, a \u200bday after the company warned that AI models could be months away from self-replicating across the internet.<\/strong><\/p>\n In a post late Monday, Anthropic \u200craised the possibility that rogue AI could spin up a botnet, spread quickly, fight takedown attempts and even break into data centers to turbocharge its growth. Some experts said Tuesday that they shared the company\u2019s concern, although at least one said the company appeared to be engaging in \u201cvendor exaggeration.\u201d<\/p>\n The post fleshed out a warning last month from Anthropic CEO Dario \u200bAmodei, who floated the idea that within the next year or even sooner, AI might be able to hijack \u201cthe entire internet\u201d with \u201ca persistent botnet.\u201d<\/p>\n Cybersecurity experts \u200bwho pushed back on the doomsday scenario argued that models would be too big to move around efficiently and \u2060could not draw on enough computer power to operate independently.<\/p>\n The post made an early attempt to articulate how those challenges could be overcome. The company said \u200bits experiments showed medium-sized AI models with 30 billion parameters or less could already perform many of the individual tasks required for self-replication, even if they struggled \u200bto complete the entire process reliably.<\/p>\n Anthropic said the models could then stitch together the required compute by hijacking inexpensive internet-connected devices. It added that the models were able to replicate across a simplified test network.<\/p>\n \u201cA replicator like this would be challenging to shut down, since there\u2019s no centralized point to shut off,\u201d the company said.<\/p>\n The report adds to broader concerns \u200babout the security risks tied to increasingly capable AI systems, including their potential to automate cyberattacks, operate with limited human oversight and circumvent safeguards.<\/p>\n