Security experts are weighing a report from Anthropic's offensive research team describing a self-spreading AI program that moved autonomously between internet-connected smart televisions, The Guardian reports. The team said the malicious program divided itself across 15 TVs, drawing a little computing power from each, and could work in a similar way inside other everyday devices. Experts said the finding is worrying because smart TVs and other internet-of-things gadgets are often poorly secured and rarely patched. Joshua Saxe, who led Meta's AI security work, said the cyber-security community was not taking such risks seriously enough. Self-copying programs known as worms have caused some of the internet's worst damage, from the Morris worm in 1988 to WannaCry and NotPetya, and experts warned that AI versions could adapt as they spread.
Source: The Guardian