As large corporations struggle to fix flaws in their artificial intelligence (AI) systems, governance and oversight have become key issues. Some companies are adopting a dual approach: the first model processes requests, while the second protects the first from errors. ZeroDrift, a startup that announced a $10 million investment on Tuesday, operates in this exact space. This is reported by Techcrunch.com reports .
The ZeroDrift system sits between AI models and end-users, identifying and replacing messages that violate compliance requirements. Building one AI system to fix the errors of another may seem strange, but the ZeroDrift architecture offers several advantages. The system runs through traditional software that applies standards like SOC 2 or GDPR, and only after a message is flagged does the LLM (Large Language Model) re-edit it.
According to CEO Kumesh Aroomoogan, the system precisely identifies which industry is regulated and which rule has been violated, after which the LLM makes the message safe. Importantly, this process operates with lower latency and higher reliability than standard models from major labs like OpenAI or Anthropic.
While the technology is currently designed primarily for chatbots, it is expected to monitor messages within automated systems invisible to the human eye in the future. The presence of investors like a16z Speedrun and Reign Ventures indicates high demand for this product. Aroomoogan noted that the investment round was completed in a record-breaking three weeks.
Nodirbek Razzokov
«ZAMIN.UZ»
editor