Why Companies Should Be Cautious About OpenAI’s New AI Browser
420OpenAI just released a new AI-powered browser called Atlas, but experts say companies should hold off on using it right now. It’s a cool tool, but it comes with some security risks that could make businesses hesitant to adopt it widely.
Atlas is only available for Mac computers for now. It’s designed to automate web browsing based on what the user wants. The browser starts with ChatGPT as its main page, and it can take over browsing tasks. For example, if someone is looking at a recipe and wants to buy ingredients, Atlas can automatically go to shopping sites like Instacart and place the order. It also has a side panel chatbot that can make changes to documents in Google Docs — which already has some automation features.
However, critics warn that Atlas could be vulnerable to “prompt hijacking.” That’s when malicious code embedded in websites tricks the AI into doing harmful things, like stealing data. The browser relies heavily on personal information to tailor browsing, which raises concerns about identity theft and intellectual property theft if the system is compromised.
According to Bob O’Donnell, a tech analyst, many companies will likely restrict access to Atlas until these security issues are addressed. “Enterprise users are already worried about security, so they’re probably going to limit or block access,” he says. Security expert Oded Vanunu also recommends that organizations treat AI browsers like Atlas as high-risk tools. They need strict monitoring, clear rules about how they can be used, and restrictions on accessing sensitive data until security improves.
Experts like Jack Gold from J. Gold Research say that companies shouldn’t deploy new browsers like Atlas without thorough testing. They need to make sure there are no security flaws and that existing apps still work properly. While Atlas offers promise for boosting productivity by automating complex web tasks and coordinating actions across multiple tabs, it’s still early days. The browser has not yet been proven safe in large, regulated environments, and its resilience to abuse or malicious prompts is untested.
Another thing to consider is that Atlas is only available on Mac right now. This limits its usefulness for many organizations that use Windows or other platforms. Experts suggest treating Atlas as a pilot project rather than replacing existing browsers until more platforms are supported and security features are strengthened.
OpenAI’s security chief, Dane Stuckey, acknowledged these concerns on social media. He warned that attackers could manipulate the AI into leaking private information or making mistakes, like buying the wrong product. Still, he says OpenAI has taken steps to improve security. The AI can detect when pages contain personal data and can operate in a “logged out” mode that doesn’t access user credentials.
Historically, OpenAI has released new versions of its tools and then fixed issues based on user feedback. Many tech giants like Google, Meta, and Netflix follow similar practices. Despite the risks, analysts see potential in Atlas, especially if it develops the ability to run AI models locally on devices without needing an internet connection. O’Donnell points out that having an on-device AI would let many people download and use advanced models without exposing data to the cloud.
Alongside Atlas, OpenAI is working on other productivity tools to compete with Microsoft 365 and Google Workspace, both of which are adding more AI features. These developments could turn browsers into central platforms for corporate and personal apps. However, the big question remains: how much exposure is OpenAI willing to give its main cloud-based ChatGPT model? Currently, full access is only available through a subscription, which might limit widespread enterprise use.
Early adopters are curious about Atlas, but many experts doubt it will quickly overtake established browsers like Chrome or Edge. Patrick Moorhead of Moor Insights and Strategy says most users and companies will probably wait until more features are built into their favorite browsers. Many already have similar AI capabilities, so Atlas needs to do a lot to stand out.
Beyond Atlas, other AI browsers are entering the scene. Perplexity just launched an AI browser called Comet, offering similar features. Meanwhile, Atlassian has acquired The Browser Co. and is developing a more secure, enterprise-focused AI browser called Dia. This move shows that the industry is taking security seriously as they explore new ways for AI to enhance web browsing.
While AI browsers promise exciting new ways to surf the web and automate tasks, security remains a major hurdle. Companies need to weigh the benefits against the risks and proceed cautiously until these tools become more mature and secure.
What do you think?
It is nice to know your opinion. Leave a comment.