It appears agentic AI is all the rage: OpenAI's new agents, "Dots," rolled out this week, during Meta's Muse just passed five myriad downloads. Rather than fair chatting alongside or using them to create images, AI companies desire users to commencement asking their bots to do things on their behalf, from managing their inboxes to ordering dinner.
But agents aren't fair the next logical stage for AI as a technology. When implemented poorly, they can be wildly dangerous. AI agents are designed to run on their own, and, in command to do that, they frequently need near-total admission to your devices and their data. Some group are now giving agents complete use of their computers, including all of their messages, files, passwords, and accounts. That's a lot of trust, particularly whenever it's not that difficult to "hack" one of these agents for malicious purposes alongside a poisoned immediate hidden on a website, which could guide your AI delegate to hand hackers the keys to your financial institution accounts. That's no good.
Apple is making it much additional difficult for you to put yourself in this circumstance
How does Apple awareness concerning this? The business is famously rearward on the application flank of AI—it lone rolled out its AI-powered type of Siri in September—but its AI hardware endeavor is booming. When the AI delegate craze archetypal took off with projects akin Moltbot, customers bought Mac minis in droves in command to run those programs. Now, Apple advertises its assorted Macs alongside AI at the forefront: It doesn't recommendation its own models akin OpenAI or Anthropic, but it's making plentifulness of prosperity marketing you the hardware to run those models.
Perhaps that's why Apple seems fairly concerned concerning the AI delegate trend, particularly whenever it comes to the data users are exposing to run these programs. Apple is all concerning privacy and security, and as such, I can lone ideate the business observing in fear as so many users willingly hand their Macs (and all of their data) complete to automated AI bots.
Against that backdrop, the business posted an update on its developer forum entitled "Updates to Full Disk Access in macOS." In it, the business highlights the dangers of developers asking users to aid "Full Disk Access" permissions to an app. Apple says this function has been traditionally reserved for backup apps that need to admission the complete disk to function. (After all, if your backup app can't admission your complete Mac, how can it rear up your complete Mac?) But now, Apple says developers are asking users to disclose everything on their Macs, including messages, mail, browsing history, and more, without those users completely understanding the consequences. Apple equal argues the privacy implications broaden beyond the user, as communication apps that petition Full Disk Access colony the group the person is chatting with, too.
And Apple appears to be doing item concerning it: While particulars are slim, the business says it volition current "additional controls to justify that users who genuinely desire to aid an app this bonzer flat of admission can lone do so alongside extremely definitive person action." I'm guessing this method that in the future, formerly you download an delegate to your Mac, you'll have to click through a sequence of warnings and settings to "prove" to macOS that you really desire to disclose your complete Mac to the app and its AI. The business says it's doing this now, as the danger volition lone develop as AI agents enhance and rotate into additional autonomous.
The risks of AI are expanding throughout the board
I'm all for additional safeguards whenever it comes to AI. In the past, you needed several flat of specialized ability to instal an AI delegate on your Mac. As such, you apt understood the risks involved, or willfully ignored all the warnings in command to do so. Now that operating agents is as uncomplicated as installing Muse on your Mac, the obstacle to admission is minimal. Anyone can do it, equal if they don't completely comprehend what they are really doing. As the AI industry reckons alongside the risks its innovation poses to humanity, companies akin Apple should certainly be figuring out how to defend their users from AI's current dangers—even if users can motionless choose to disregard those warnings.
Covering tech news, how-to guides, explainers, and more.
