1000+ unique media and news posts every 24 hours…
Apple released the next wave of Apple Intelligence on Wednesday, which brings ChatGPT integration iOS 18.2 to supported iPhone models. As part of the 12 Days of ChatGPT event, OpenAI held a livestream to demonstrate the new features.
It shouldn’t be surprising to see Google drops the big Gemini 2 announcements on the same day. While it could be a coincidence, it certainly seems like Google wanted to steal Apple and OpenAI’s thunder. Everyone, including Google, knew that iOS 18.2 would be released this week with ChatGPT in tow.
As a long-time iPhone and ChatGPT user who still can’t access early Apple Intelligence features (I live in Europe), I’ll say I’m already jealous of Gemini 2. It’s not about the better, faster capabilities you expect from the next generation Twin models. It’s about the AI agents Googling revealed, including Project Mariner, which lets you use Gemini to surf the web and perform tasks for you.
Early reports told us that Google would unveil such a tool, which was called Jarvis at the time. I explained that I wanted AI agents to surf the web for me, but Google Jarvis wouldn’t be it.
Now that Project Zeeman is officially the name of Google’s Chrome browsing AI, and we know what it can do and how it behaves, I’ll say it looks great. There’s nothing like it in ChatGPT, and it will be a while before Siri can surf the web for me.
My privacy concerns persist, however, because Google hasn’t provided enough details about what happens to my data once I tell Project Mariner to surf the web for me.
Project Mariner is built on top of Gemini 2. It can surf the web for you and read the information on the screen, including pixels, text, code, images and forms. It can then accept commands and perform tasks for you. Simply type the prompts into a Chrome extension and Project Mariner will get started.
The Gemini 2 AI agent can surf the web for you, but only if you keep the Chrome tab open and focused. That’s a good thing, because it keeps Project Mariner from surfing the Internet in the background without your knowledge.
The disadvantage is that you cannot (yet) work in another window, which is what I want from such AI agents. Imagine telling ChatGPT or Gemini to research topics in the background while you work on something else.
Back to Project Mariner, the AI agent will type, scroll and click on that active tab to complete the tasks in the prompt. It can use reasoning to determine the steps needed to perform its task.
You’ll also see a log of steps performed, so you always know what Project Mariner is doing. You can stop it if you need to. Google said in a blog post that Mariner will also ask for confirmation before taking sensitive actions, such as purchasing something on your behalf.
In the video above, the user asks Project Mariner to look up email addresses for a list of companies in a spreadsheet. That’s a good example of what AI agents will be good for. We do this tedious research work all the time while surfing the Internet. It is also time-consuming because we have to visit different websites, scroll, click and save information.
The Project Mariner demo proves that. It takes approximately 12 minutes to complete the task. Google sped up the video for the demo, but didn’t recreate it like in the early days of the Gemini demos.
Google does think about safety also, and says it’s building tools into Project Mariner to ignore hacker attacks:
With Project Mariner, we are working to ensure that the model learns to prioritize user instructions over third-party attempts at quick injection, so that it can identify potentially malicious instructions from external sources and prevent exploitation. This prevents users from being exposed to fraud and phishing attempts, for example through malicious instructions hidden in emails, documents or websites.
However, Google has never said anything about user privacy when it comes to Project Mariner.
We may still be in the early days of AI agents, but Google needs to make the privacy implications clear to the user from the start. It should follow the lead of Apple, which did just that when unveiling Apple Intelligence at WWDC. Apple has made sure we understand how AI privacy works, building strong privacy protections into Apple Intelligence that includes both on-device and cloud AI processing.
With Project Mariner I have no idea what happens to my data. It’s not just about my prompt training Gemini, which I would like to opt out of from the start. It’s about how Project Mariner works.
I would like the AI agent to work on the computer instead of in Google’s cloud, to minimize data exchange between the browser and Google’s server. After all, all he has to do for me is surf the Internet. The PC must be powerful enough to process data on the device.
If not, Google should explain how it protects the data that reaches its servers and what it does with it.
Since the AI would browse websites for me, I’d also like to be able to customize how Mariner works. I would like the AI agent to always reject cookies when opening a new page. I also wouldn’t want Google to use AI agents’ browsing data to improve the profile they have about me for advertising purposes.
Put another way, I don’t want to pay for AI agents with my data, whether it’s Project Mariner or a future AI agent from OpenAI or Apple. I’d rather it be a premium feature that costs money.
Finally, I will also discuss the elephants in Google Chrome and Google Search. Project Mariner runs in Chrome and searches the web via Google Search. I wouldn’t want any of that. The video above is the full demo of the previous clip. This allows you to see exactly how Mariner works in Chrome and how Google Search is used to find the websites.
Let’s say I end up using Project Mariner; I want to be able to use the extension in another browser. Furthermore, Project Mariner should use whatever default search engine I rely on, which won’t be Google’s.
That may seem like wishful thinking at this point. Google wants Gemini 2.0 built into several apps and services, including Search, Maps, Docs, and others. That’s obviously great for people who want to get the most out of Google’s ecosystem. It also gives Google a competitive advantage over competitors.
Remember that though Google lost a major antitrust case targeting Google Search. The company is a monopolist in the US. It wouldn’t look good if Project Mariner only worked in Chrome and only with Google Search. But if Google sells Chrome to settle that antitrust lawsuit, that will be a different matter.
Finally, I’d also like to say that Google can always claim that the same privacy rules it established for Gemini will apply to Project Mariner, Astra, Jules, and any other AI agents it works on. But that’s not good enough. We really need to know what’s happening to our data in chats with the AI agent from the moment we send the prompts to the moment we get answers.
1000+ unique media and news posts every 24 hours…