Building Content Safety into Burgundy, a Blog Writing Copilot

I recently spoke at TAICO - Toronto AI and Cybersecurity Meetup about “Building Content Safety into Burgundy, a Blog Writing Copilot.”

I discussed our journey with Burgundy, the motivation for customizable safety features, and the limitations of existing methods like OpenAI Moderation and LlamaGuard. The session included a live demo of our safety implementation and methodology.

The slides from the presentation are here. The relevant code snippets will be shared in August. Enjoy and feedback, thoughts are very welcome!