Inspiration
Most people blindly accept Terms & Conditions without reading or understanding them, often loss of rights. We wanted to create a tool that brings transparency and critical awareness to these dense legal documents.
What it does
TrustCheck scans any TOS or privacy policy, flags manipulative or risky clauses, gives each clause a simple summary, and rates the overall "trustworthiness" of the agreement. It highlights dark patterns and shows key user risks in plain language.
How we built it
We used Python with Flask for the backend. The documents are scraped with open-source libraries and split into clauses. Each clause is analyzed by a large language model (using DeepSeek API), and we display results via a Chrome extension window with color codes and easy summaries.
Challenges we ran into
Splitting complex legal text into meaningful clauses was harder than expected, and tuning the language model to avoid vague or unstructured outputs required experimenting with prompts.
Accomplishments that we're proud of
We delivered a working demo that can reveal real risks and "gotchas" in widely used agreements, making legalese accessible for non-experts.
What we learned
Prompt design and data chunking are crucial when using AI for regulatory or legal analysis. Usability matters just as much as technical accuracy.
What's next for TrustCheck
In the future, TrustCheck will use collected data to train models that deliver more accurate and personalized responses for each user, while also expanding support to more languages. We also want to optimize the mobile app connectivity.
Log in or sign up for Devpost to join the conversation.