← Back to Model Beat
Policy·Aug 22·all news from August 22, 2026

Presentation: SafeChat: Building AI-Powered Safety Systems at Scale in a Real-Time Marketplace

DoorDash has developed SafeChat, an AI moderation platform that balances high-speed internal models with LLM analysis to filter harmful content in real-time. By moving away from reliance on expensive, standalone LLM pipelines, the company reduced moderation costs while maintaining the ability to handle complex, nuanced safety decisions.

Covered by 1 source

Related stories

PolicyBill Gates warns AI is more dangerous than the tech industry will admitAug 26 · 38 sourcesPolicyDisrupting a new covert influence campaign from RussiaAug 25 · 4 sourcesPolicyFrom Preferences to Principles: Rubric-Based Alignment for Grounded Knowledge AnswersAug 26 · 2 sourcesPolicyAnthropic Plans to Change Data Retention Policy for Advanced AIAug 20 · 2 sources