OpenAI has announced the launch of its AI-powered browser, ChatGPT Atlas, aimed at competing with Google for online information searches. The browser will initially be available on macOS, with plans to expand to Windows, iOS, and Android, and it will be accessible to all free users at launch. Ben Goodger, Engineering Lead for ChatGPT Atlas, emphasized that the product integrates core ChatGPT functionalities, allowing users to interact with search results directly. This includes a sidecar feature that enhances user experience by eliminating the need to copy and paste information into the chatbot, thus streamlining online tasks. While existing AI-powered browsers have shown promise, OpenAI’s ChatGPT Atlas aims to refine these capabilities further, particularly by allowing users to automate web-based tasks through an agent mode, although this feature will only be available to Plus, Pro, and Business tier users at launch. The effectiveness of this new browser in challenging Google Chrome’s dominance, which boasts over 3 billion users, remains to be seen.
Simon Willison took it for a spin, and noted one of the standout features of ChatGPT Atlas is its “browser memories,” which enables the AI to remember details from users’ browsing sessions to enhance interaction and provide personalized suggestions. However, concerns regarding security and privacy have been raised, particularly about the experimental “agent mode” that allows ChatGPT to navigate and perform tasks on behalf of users. Users must remain vigilant, as the current safeguards may not fully prevent risks associated with prompt injection and other potential vulnerabilities. Website developers can also enhance interactions by implementing ARIA tags, which improve how the AI interprets page structures and elements, making it easier for users to engage with their sites effectively.
ChatGPT is processing an astonishing 2.5 billion messages daily, according to recent data from OpenAI. This rapid adoption of artificial intelligence has seen the chatbot’s user base expand to approximately 800 million, which represents over 10% of the adult population worldwide. Research indicates that 70% of the queries directed to ChatGPT are not work-related, with common requests falling into categories such as practical guidance, information gathering, and creative writing. Notably, OpenAI has strengthened its partnerships with major tech companies and is branching out into consumer industries, as evidenced by a recent collaboration with Walmart allowing users to purchase items directly through the chatbot.
The Register noted that OpenAI is reportedly facing significant financial challenges, losing approximately three times more money than it earns, with 95 percent of its ChatGPT users not paying for the service. In the first half of 2025, the company generated $4.3 billion in revenue but suffered a net loss of $13.5 billion, as detailed in a report by The Information. Notably, about 70 percent of OpenAI’s recurring revenue derives from paid ChatGPT subscriptions, yet only 5 percent of its 800 million users contribute financially. Despite this disparity, venture capital firm Menlo Ventures noted that this gap presents a substantial opportunity for growth in the market. OpenAI’s CEO, Sam Altman, has set ambitious goals to double the customer base while investing heavily in data center capacity, aiming to secure more than 26 gigawatts from major tech companies, including Nvidia and Oracle, at a projected cost exceeding $1 trillion.
And some context about the web itself. A recent report reveals that over half of new articles published online are now generated by artificial intelligence. According to SEO firm Graphite, this trend saw AI-generated content slightly surpass human-written articles between November 2024 and March 2025, peaking at 55% in January 2025. Despite this rise, the share of AI-generated articles appears to be stabilizing, as they tend not to perform well in search results, raising concerns about their effectiveness as reliable information sources. A survey from the Pew Research Center shows that only 2% of Americans regularly obtain news from AI, with the majority expressing distrust in such content.
Why do we care?
So OpenAI’s decided to make a browser—ChatGPT Atlas—and it’s basically Chrome with an AI brain. You can search, chat, and even have it do stuff on the web for you if you’re a paying user. Sounds slick, but let’s not miss the big picture: OpenAI’s burning through cash—losing three dollars for every one it earns—and it’s building this trillion-dollar data center empire to feed the model.
Here’s the kicker: the web Atlas will crawl is already more than half AI-generated, and people barely trust it. So OpenAI’s browser is reading content written by AI, summarizing it with AI, and serving it to humans who aren’t sure they want any of it. That’s a snake eating its own tail.
If you’re running an MSP, this is your early warning: agent-mode browsers are coming, and they’ll blur the line between user and bot. Lock down your browser policies, figure out where automation makes sense, and make sure data doesn’t leak out through these new tools.
Atlas might not beat Chrome, but it will change how people interact with the web—and that’s something IT providers can’t afford to ignore.