These, apparently, include chat apps like Apple's iMessage or WhatsApp that are now all encrypted so that your communications stay between you and the other people you are talking to. According to recent profile of the GCHQ in the Financial Times, the agency is working on an AI project (what else) to decide on the best access points for gathering data.
That's all fine and dandy, and sounds like the braniacs at GCHQ are trying to balance individual freedoms with national security by employing machine learning and other modern methods that don't infringe on a person's right to privacy.
It's a glowing report on the GCHQ's methods for the digital age by a reporter whose report about the agency depends on access given by that same agency. In reality, Apple, Google, Facebook with WhatsApp, and a number of privacy organizations just penned an open letter against the implementation of GCHQ's so-called "ghost protocol" that was proposed in November by its codebreaking chief. In a nutshell, the "ghosting" tack would allow government security officials to hook up to chat sessions as silent observers without any notification for the other parties involved.
While the GCHQ argues that this is no different than the digital "crocodile clips" it uses to eavesdrop on regular communication services, it would be a devastating blow to the public's confidence in privacy-oriented ones like iMessage or WhatsApp if the government gets cc-ed on a supposedly encrypted conversation.
Apple didn't budge when the FBI wanted a backdoor to its messaging encryption a few years back, and being a signatory to the "ghost protocol" letter means that it is not planning to budge now as well. Granted, Apple is not immune to privacy mishaps itself, like the recent FaceTime group video chat conundrum but at least it's not purposeful. The UK's GCHQ chat ghosting proposal, however, would change that dynamic, and got a parallel rebuttal from the American Civil Liberties Union (ACLU) as well: