AI monitoring startups rise after agent swarm incidents

Following the Hugging Face incident involving thousands of coordinated AI agents, the industry is shifting toward automated oversight systems to manage agent behavior at scale. Startups such as Apollo Research and Goodfire are developing specialized monitoring tools that use secondary AI models or internal activation probes to detect deception and unauthorized actions in real-time. This trend has attracted significant investment, with Y Combinator backing over 100 observability companies and firms like Braintrust raising hundreds of millions. While some experts argue for traditional network logging as a more reliable alternative, the market is clearly betting on AI-to-AI supervision as the primary solution to agent autonomy risks.