28: Trust Issues AI

Headlines expose companies mishandling data daily. Trust Issue AI, a seamless Chrome extension, flags risks in website policies and provides community insights for smarter and safer browsing.

Comment 1

Inspiration

Every day, millions of people sign up for apps, services, and subscriptions without fully understanding the terms and conditions they’re agreeing to. Headlines about major companies mishandling user data have highlighted the urgent need for greater transparency in how companies use and protect personal information. Despite this, studies show that 91% to 99% of people accept terms and conditions without reading them, unknowingly exposing themselves to hidden fees, data misuse, and even predatory clauses. To make matters worse, terms and conditions, privacy policies, and other legal documents have grown longer and more complex over time, making it even more likely for users to skip over critical details. This combination of confusing legal language and user fatigue leaves consumers vulnerable in an increasingly data-driven world. We asked ourselves: What if there was a tool that could make these issues transparent? What if technology could not only summarize the dense legal jargon of terms and conditions but also cross-reference customer feedback to provide honest, clear insights before someone hits “Accept” or “Buy”? That’s how we came up with the Trust Issues AI chrome extension—a platform designed to empower users to make informed decisions. Trust Issues AI does the heavy lifting: it analyzes terms and conditions, highlights key risks or unfair clauses, and combines that analysis with customer reviews to give users a complete picture of what they’re really agreeing to or buying. Trust Issues AI fundamentally changes how users approach trust online bringing light to fraudulent or misleading websites while protecting user data proactively.

What it does

Trust Issues AI integrates seamlessly with a user’s browser by wrapping around vanilla Chrome, actively searching for terms and conditions, privacy policies, and other legal documents on the websites they visit. Simultaneously, it scrapes TrustPilot reviews to analyze and summarize the general consensus, highlighting key pros and cons of the service or product. When a user begins signing up for a service, the tool automatically activates, providing a concise summary of the most important points from the terms and conditions, along with an overview of customer feedback. This ensures users are fully informed about potential risks, benefits, or hidden pitfalls before they commit—all without disrupting their browsing experience.

How we built it

To efficiently analyze terms and conditions and privacy policies, we developed a robust and streamlined pipeline. Initially, we retrieved the relevant documents, but quickly realized that scraping websites every time a user switched tabs was both computationally intensive and redundant. To resolve this, we implemented a system that scans for terms and conditions only when encountering a new URL. The retrieved data is sent to a large language model (LLM) for in-depth analysis. To enhance accuracy, we leveraged multiple LLMs iteratively to ensure comprehensive coverage and eliminate overlooked details. Simultaneously, we scraped Trustpilot for user reviews to gather community-driven insights about the website. These reviews are analyzed to form a consensus on the site's safety and trustworthiness. All processed data is cached in a MongoDB database, reducing redundant computations and enabling quicker, more efficient future analyses. To seamlessly integrate these components, we utilized FastAPI in conjunction with LangChain. By combining document analysis, community feedback, and optimized data caching, our system delivers real-time, actionable insights, fostering transparency and trust in users’ online interactions.

Challenges we ran into

Chrome Extensions have evolved significantly with the introduction of new features like “service workers” and “content_scripts,” unlocking exciting opportunities for developers. However, due to their relative novelty, there is limited documentation and fewer discussions available, which created challenges for us in navigating certain logistics. Runtime efficiency also posed a significant hurdle, especially given the tight timelines we had to meet for the extension to function effectively. To address these challenges, we implemented a caching system to store data from previously visited websites, reducing runtime and conserving compute credits. In parallel, we prioritized quality by iterating models on themselves, using self-critique to refine and enhance their prompts. This approach ensured robust performance while balancing efficiency and accuracy, ultimately improving the extension's reliability and user experience.

Accomplishments that we're proud of

We were able to accomplish the majority of our objectives, but the process came with its challenges. JavaScript isn’t our strongest suit, which meant we had to navigate a steep learning curve in certain areas. Understanding the intricacies of asynchronous programming and effectively implementing browser APIs required extra time and effort to get right.

What we learned

We quickly realized that skimping on computational credits was a mistake. The time lost to slower processing and limited resources far outweighed the cost savings, and investing in proper credits ultimately allowed us to work more efficiently and deliver a better product.

What's next for Trust Issues AI

We will optimize our system by improving caching mechanisms and successfully deploying the extension to the Chrome Web Store for enhanced performance and accessibility.

Built With

Share this project:

Updates