The leading artificial intelligence company Anthropic had previously warned of the ability of AI models to replicate themselves inside smart home devices, noting in the study it published that this could cause malicious models to replicate themselves inside these devices to escape human control, which makes it more difficult to stop them from working.
Hot Topics
the Red Team within Anthropic, the name the company internally gives to its security testing team, as saying that they had successfully tested this model and made it divide itself into small parts distributed across 15 smart TVs connected to the same internet network.
Joshua Sachs, former head of AI security development efforts at Meta, explained that Anthropic's latest discovery is just as worrying as any other wave of cyberattacks, adding, "I sometimes worry that people in the cybersecurity community aren't taking these very real dangers seriously enough."
this technology as "AI worms," as these worms can infiltrate smart homes and then reproduce inside them without the owner's knowledge and without any real control over them.
But what harm could such worms cause? And do they really warrant all this fear?
Reproduction without human intervention
Typically, malware needs to trick a person in order to infect their devices and spread within them. When a user clicks on a malicious link, the software spreads within their computer or phone to access their various data and accounts.
However, the situation is different with one category of malware to which these worms belong; the idea of self-replicating worms is not new to cybersecurity experts.
Malicious worms have previously caused damage to internet networks and users' devices, described in a Guardian report as "the worst in internet history," clearly indicating that such crises have occurred before.
A legacy of disastrous breaches
The legacy of malicious worms includes a number of dangerous cyberattacks dating back to the period before the internet became widespread among the general public.
Tags:
technology