Illusion

Inspiration

Users have very little clarity on how companies collect, use, share, and retain their data. Key information is often scattered across multiple pages, buried in obscure links, or written in long, vague, and inconsistent policies that are hard to interpret. As a result, people struggle to make informed choices about the services they use, exposing them to hidden privacy and data-tracking risks.

What it does

With the user entering a company name or URL, we report back important information which helps them understand about the corporates. The product creates a standardised, objective metric for privacy, potentially forcing companies worldwide to adopt better practices to compete for high scores, similar to how environmental or credit scores operate.

• Risk ratings – evidence-based assessments of data collection, sharing, and tracking risks.
• Transparency scores – how open and clear a company’s policies actually are.
• Summary of risks – a concise overview of the major findings.
• Protection advice – guidance on how users can safeguard their data.
• Source citations – verifiable links supporting every claim and decision.
• Transparency – reveal the token usage, total cost, processing time and tool used, alongside with the thought process. ## How we built it We built a transparent multi-agent system that automatically analyses a company’s data practices.
• Search Agent - finds official policy source from the company website.
• Extraction Agent - pulls only the relevant information.
• Summary Agent - produces a clear, source-backed transparency report. The entire process is fully auditable, with logged reasoning steps, tool calls, and trace metadata to ensure reliability and robustness. ## Challenges we ran into
• Integrating Amazon bedrock into Strands-agents.
• Problem with caching the data. ## Accomplishments that we're proud of
• Integrating Langsmith into the Strands-agents.
• Successfully combining multiple agents in an efficient manner. ## What we learned
• Effective use of Holistic AI.
• ReAct design pattern in AI.
• Effective collaboration by delegating important task to team members. ## What's next for Illusion Comparing multiple company data, transparency level and give sensible suggestions to the user. The suggestions will be justified and have clear reasoning, showing the sources of information.

Built With

• amazon-web-services
• fast-api
• gemini
• javascript
• langsmith
• mlflow
• python
• react
• sql
• strands
• strands-agents
• strands-agents-tools
• typescript