Chrome's agentic capabilities powered by Gemini to arrive in December with v143, giving OpenAI's Atlas a head start

Google just handed OpenAI a pretty sweet window of opportunity. While ChatGPT Atlas launched on October 21 with its agent features already baked in, Chrome users waiting for similar capabilities might be stuck twiddling their thumbs until December 2. Google’s official support documentation confirms that agentic capabilities will go live with v143. And Chromium’s schedule indicates that v143 will arrive in early December.

That’s over a month where OpenAI gets to court Chrome’s massive user base without any real competition from Google itself.

Remember when Google announced Gemini in Chrome last month? The hype was real, but the actual product felt pretty underwhelming. Right now, it’s basically just Gemini living in your sidebar. You can ask it to explain what’s on your current page or fire off random questions without opening a new tab, but that’s about it. The good stuff – the part where Gemini books your haircut appointment or handles your grocery order while you do literally anything else – was promised for “the coming months”.

So yeah, “the coming months” means December. The official Chromium release schedule shows Chrome 143 getting its beta promotion on October 29, followed by a stable cut on November 18, before finally landing in users’ hands on December 2. OpenAI’s Atlas doesn’t have to wait for any of that.

And it’s not just Atlas nipping at Chrome’s heels. Perplexity dropped the paywall on its Comet browser earlier this month, so anyone can jump in and try AI-powered browsing right now. Both browsers have that killer feature where the AI assistant can just perform actions on websites without you having to lift a finger like some kind of caveman.

Here’s the million-dollar question though: does it even matter? Chrome absolutely dominates browser market share. But if people actually want agentic features today, they’ve got options that don’t involve waiting around for Google to ship an update. Will those early adopters stick with Atlas or Comet once Chrome catches up? Or will Chrome’s install base make this whole head start irrelevant?​

There’s a darker side to this race that nobody’s really talking about enough. Brave just published research on what they’re calling “unseeable prompt injections,” and honestly, it’s pretty terrifying. The basic problem is that AI browsers can’t really tell the difference between what you’re telling them to do and what a sketchy website is secretly telling them to do.​

So, let’s say, for example, you visit a website that has invisible text (white text on white background, that kind of thing) with instructions hidden in it. Your AI browser reads that text and thinks those instructions came from you. Now imagine you’re using that browser to book a flight or buy something online, and a malicious site redirects those actions to drain your bank account instead. Brave found these exact vulnerabilities in Comet, and the security concerns apply to pretty much every agentic browser out there.

Maybe Google’s delay isn’t such a bad thing after all. They’ve got time to figure out better protections against these prompt injection attacks before exposing hundreds of millions of users to the risk. Then again, maybe OpenAI and Perplexity grab enough market share in the meantime that even a more secure Chrome struggles to win people back.

The next month or so should tell us whether being first actually matters, or if Chrome’s dominance is too big to dent. That said, do watch out for Gemini 3.0, which is also expected to drop around the same time as these agentic capabilities. Chatter online suggests it might just be the best AI model out there, and with it possibly powering Chrome’s agentic capabilities, maybe it’ll also offer the best user experience.

Look, there’s always a chance that things don’t go according to plan, and the agentic capabilities don’t make it to users in December. But right now, that’s all the information we have to work on.

Now we wait…