This week, OpenAI gave us a sneak peek into its latest creation: designing a web browser d ChatGPT Atlas. The idea? To enhance your online surfing experience using artificial intelligence. But here’s the kicker: your permission allows ChatGPT to track and remember almost every move you make online.
Generously surpassing Google Chrome in surveillance capabilities, Atlas keeps tabs not only on which sites you frequent but goes a step further by logging your online activities and preferences while on those sites. Oh, and it can even take the reins of your mouse to browse for you. How convenient, right?
While it’s still early days for Atlas and we can’t fully judge its AI tricks, there’s a massive gray area regarding user privacy that raises eyebrows. Navigating how it manages what it recalls about you might leave you feeling a tad puzzled.
Your web browser choice is bigger than you might think. It serves as your main gateway to the digital world — a treasure trove where companies can gather info on you to serve ads, direct you to specific locations, and even train their AI systems based on your behaviors.
OpenAI isn’t flying solo in the quest of integrating AI into browsing. We also have Perplexity with its Comet browser, and Google is incorporating features from its Gemini bot into Chrome, sporting capabilities that enable AI to perform tasks for users (the Washington Post collaborates with both OpenAI and Perplexity, just so you know).
So, what exactly can Atlas do? It shifts the spotlight away from Google and places ChatGPT at the helm for discovering websites and content. There’s an “Ask ChatGPT” button perched in the corner, enabling you to engage in real-time conversations with the bot regarding the pages you’re browsing. Want a summary of an article? It can handle that. Needing assistance revising those pesky email drafts? Just one click away!
Product lead Adam Fry from OpenAI echoed that the primary goal was about teaming up AI with users in their web journey.
However, a closer inspection of Atlas’s privacy and data management practices might leave you uneasy. If you choose to authorize it, this browser builds a vault of memory about your visited sites, drip-feeding snippets back at just the right moment. You could easily tell Atlas: “Reopen the Halloween stuff I was looking at in some tabs last week,” and it’s ready to comply.
More than just URLs, Atlas memorizes snippets tagged as “facts and insights” that are essentially summaries that OpenAI keeps stored on its servers. According to Fry, it might recognize that you have upcoming travel, have a favorite airline in Delta, and use Google Calendar.
While this deeper level of personalization has its benefits, it also invites complexity regarding your privacy, making it crucial to control Sho-assort memories that you’d prefer not be recalled later, like awkward relationship inquiries or late-night health condition searches.
When it comes to data retention specifics, things can turn murky fast. OpenAI suggests it’s all about learning tasks and preferences minus sensitive data like government IDs, banking info, home addresses and so on. Though, a test cited that Atlas did save recollections* details related to health service registrations. Lena Cohen from the Electronic Frontier Foundation voiced concerns about potential privacy issues stemming from this extensive information gathering. OpenAI claimed to have upgraded their systems post-reporting on this, though.
Yes, you can exert some sort of control over Atlas’s memories, but it takes work. Instructions to suppress the recall of specific sites can be given through the address bar, and cleaning your browsing history is expected to eradicate memories linked to that timeline. The settings showcase a comprehensive rundown of its archived memories, with options for deleting them either individually or en masse. Note this memory system operates separately from the running log ChatGPT has built about you.
There’s an “incognito” mode offered too, ensuring that your browsing acts without creating history. Nonetheless, much like Chrome’s similar mode, it won’t keep you off radar from other sites or even from ChatGPT.
Web companies love publicizing that they offer users autonomy in data gathering, but scattering privacy preferences doesn’t equate to genuine control. A jumbo jet might boast an array of controls, but that certainly doesn’t mean piecing them together makes one a pilot.
Fry remarked that the aim behind the memory function was to fortify Atlas’s features—not to juggle data the way social media platforms do.
OpenAI also pointed out they have no current ambition in the advertising arena.
Concerns linger about what online monitoring could mean in real-life scenarios. Let’s say governmental bodies …………..ابيع events they might find questionable? Following an update on this narrative, OpenAI informed me they would quell data sharing under “valid” legal stipulations and in emergencies. Additionally, any memorized data held by the browser clears off after 30 days and is wiped entirely.
In some respects, Atlas behaves conservatively. By default, it refrains from utilizing browsing content for AI training — there’s an option allowing you to switch this off if you wish, while ChatGPT conversations get trained unless declined outright.
Realistically, handing your browsing session to an AI introduces brand new issues, as these agents can conduct shopping events and assist with content management. Yet errors can happen, and providing them your login and payment information elevates the stakes.
In addressing safety, Fry explained that any tasks operating at higher risk are carried out in a stripped-down version of the browser, instilling alerts to supervise actions.
From the basic user experience to advanced data factors, browsers have evolved into essential collection instruments. Atlas may streamline your browsing adventure—provided you’re A-OK with handing AI detailed access to your digital life.
