ConsentReceipt

Inspiration

Most people accept privacy policies without understanding them. They are long, complex, and filled with legal language, making informed consent difficult. We built ConsentReceipt to make privacy policies understandable while ensuring users don't have to blindly trust an AI-generated summary.

What it does

ConsentReceipt transforms privacy policies into five clear disclosures: what data is collected, how it is used, who it is shared with, user rights, and data retention.

Users can paste a policy or upload a document. AI translates it into plain language, and every disclosure includes evidence from the original policy. ConsentReceipt independently verifies those quotations and flags information that is unclear or missing.

Users can also highlight evidence in the original text, translate summaries, use text-to-speech, and export their ConsentReceipt as a PDF.

How we built it

We built the backend using Python and FastAPI, with HTML, CSS, and JavaScript for the frontend. Groq powers our AI analysis and returns structured results across the five disclosure categories.

Our verification layer then searches the original policy for each AI-provided quotation instead of automatically trusting the model. Each receipt also includes verification statistics, a timestamp, and a SHA-256 fingerprint of the analyzed policy.

We also created an evaluation suite using fictional privacy policies to test citation validity, classification accuracy, misleading statements, and processing speed.

Challenges we ran into

Our biggest challenge was AI hallucination. A model can generate convincing information that isn't actually supported by the source.

We addressed this by adding deterministic evidence verification after the AI response. If evidence cannot be found, ConsentReceipt warns the user instead of presenting it as verified.

Privacy policies can also be intentionally vague, so we distinguish between information that is disclosed, unclear, or not found rather than forcing the AI to provide an answer.

Accomplishments that we're proud of

We're proud that ConsentReceipt goes beyond being another AI summarizer. Our evidence-first approach allows users to verify AI-generated claims directly against the original policy.

We also built a complete experience with document uploads, source highlighting, multilingual summaries, text-to-speech, downloadable reports, and automated evaluation.

Most importantly, ConsentReceipt follows a core Responsible AI principle: AI should help people make informed decisions without requiring them to blindly trust AI.

What we learned

We learned that responsible AI requires more than prompt engineering. Combining generative AI with deterministic verification, measurable testing, and transparent limitations creates a much more trustworthy system.

We also learned that finding a quotation doesn't guarantee the AI interpreted it correctly, which is why we keep the original evidence visible to the user.

What's next for ConsentReceipt

We want to expand ConsentReceipt into a broader privacy transparency platform that can track policy changes, compare companies' privacy practices, and eventually analyze policies before users click "Accept."

We also want to add stronger semantic verification to determine whether an AI-generated summary is actually supported by its cited evidence.

Our goal is simple: make digital consent understandable, verifiable, and accessible to everyone.