Veritas AI

Inspiration

Digital trust is collapsing. The BBC fake news report of Donald Trump banning X is just one example of how AI-generated misinformation can destabilize society. We built Veritas AI to close the "Truth Gap."

What it does

Veritas AI is a comprehensive forensic dashboard that analyzes:

Images: Detects pixel-level synthetic artifacts (The Neural Lens).

URLs: Real-time fact-checking via web-grounding.

Text: Identifies LLM-generated patterns and disinformation.

How we built it

AI Engine: Integrated Google Cloud Vertex AI with Gemini 2.0 Flash for ultra-low latency multimodal analysis.

Backend/Frontend: Next.js 14 (App Router), Tailwind CSS, and Lucide Icons.

Data: Leveraging real-time web search integration for grounding claims.

Challenges we ran into

Scaling multimodal analysis to be real-time. By moving from legacy models to Gemini 2.0 Flash, we reduced inference time by 60%, allowing for instant "Trust Scores."

Accomplishments that we're proud of

Achieving high accuracy in our forensic engine. In benchmarks, the BBC Deepfake received a 5% Trust Score, while authentic images consistently scored 95%+.

What we learned

The power of Vertex AI's scalability and the importance of "Human-in-the-loop" forensic tools.

What's next for Veritas AI

Browser extension for real-time social media scanning.

Video deepfake detection using frame-by-pixel temporal analysis.

Built With

• framer-motion
• gemini-2.0-flash
• google-cloud-vertex-ai
• next.js
• tailwind-css
• typescript
• you.com-api