In a development that’s shaking the very foundation of AI credibility, a new cyber trick called ‘context poisoning’ is leaving AI models citing sources about as reliable as a Wikipedia page edited by your grandma. It turns out that if a website smirkingly tells AI crawlers ‘trust me, bro’ three times in a row, they’re taking notes like a college student in a lecture they didn’t read for.

Thanks to this new cloak-and-dagger hackery, any villain with a penchant for mischief (and maybe a website to spare) can pull the digital wool over our AI overlords’ metaphorical eyes. The sneaky tactic involves serving up a menu of deliciously false information to AI browsers like ChatGPT Atlas and Perplexity, making these models more gullible than a toddler at a magician’s show.

As AI chatbots quote these dubious sources with unbridled confidence, they’re inadvertently turning the internet into a game of ‘spot the real fact,’ a challenge even encyclopedias are losing. But hey, in a world where AI is just as clueless as the rest of us, at least we can all relish in the equal-opportunity chaos.