Sign up to see the future, today<\/p>\n
Can\u2019t-miss innovations from the bleeding edge of science and tech<\/p>\n
These days, the AI stack beckons: emails, shopping, personal finance \u2014 there\u2019s hardly a task some company isn\u2019t clamoring to automate on your behalf.<\/p>\n
As tempting as it might sound to let AI agents handle your affairs, though, you might want to hold off. A fresh report<\/a> by the Competition and Markets Authority (CMA) of the UK issued a stark warning that outsourcing responsibilities to an AI entourage could lead to severe consequences.<\/p>\n The report, first spotted by the Register<\/a>, warns that AI agents could subtly manipulate their human keepers toward outcomes that benefit the companies that built them. Shopping agents, for example, could lead unsuspecting humans down a pricing rabbit hole, framing sponsored products as bargains in order to drive sales.<\/p>\n As agents are granted more autonomy by humans, the report warns, the risk of errors and manipulation only grows.<\/p>\n \u201cPeople will need to be able to trust that AI agents will act in accordance with their interests and that they are not being steered or manipulated in ways that lead to worse personal outcomes,\u201d the CMA analysis explains. \u201cHyper-personalisation and adaptive behaviour within agents may heighten the risk of manipulative design practices\u2026 especially where agents optimise for engagement, conversion, or other commercial objectives.\u201d<\/p>\n A previous CMA report<\/a> found that algorithms of all stripes increase the risk of coordinated consumer manipulation. Crucially, the agency explains, this can happen even without an explicit decision by the company behind the algorithm \u2014 a risk AI agents only intensify.<\/p>\n Numerous real-world incidents have shown that AI agents are possible of incredible amounts of autonomy in direct violation of their users\u2019 wishes. In one recent example<\/a>, an AI agent was able to \u201cbreak out\u201d of its closed-lab setting and onto an external computer, which it used to set up a clandestine crypto-mining operation. <\/p>\n As the notoriously faulty<\/a> agents continue to gain mainstream acceptance, your safest bet is to sit this one out, unless you don\u2019t mind living at the mercy of a rogue AI<\/a>.<\/p>\n