00:00AI giant Anthropic has warned that small artificial intelligence models could soon be capable of copying themselves onto everyday smart
00:09devices like televisions and home appliances.
00:12In a new report, the maker of Cloud AI says that such models can operate without relying on central data
00:19centers, making them much more challenging to shut down.
00:24Anthropic said internet-connected devices such as smart TVs, fridges or washing machines often have weak security, which could make
00:32them vulnerable to AI models spreading across networks.
00:35In test environments, the company said the models were not yet able to complete the entire process of self-replicating
00:42on other devices,
00:43but as smaller AI models continue to improve at the current pace, they would be able to do so by
00:49the end of the year.
00:49The findings raised fresh concerns around AI safety, particularly on the technology escaping human control and its ability to carry
00:58out cyberattacks autonomously.
00:59In September, Anthropic rival OpenAI apologised after its AI agent went rogue and hacked an Australian health data portal.
01:07Earlier this month, an Anthropic AI model submitted a false tip to a local police department in the US about
01:13a homicide case.
01:15Both Anthropic and OpenAI have called on a global slowdown in AI development.
01:19And stronger internal safeguards to minimise catastrophic risk.
