Cohere offers to put agents in lockdown mode with strict ACLs

Canadian AI model developer Cohere has unveiled an updated version of its North agent harness, Cohere North 2, to provide enterprises with a safer and more consistent experience with AI agents. The company unveiled North 2 on Monday, focusing on collaboration, access control, token spend, and overall security of the platform. Cohere North 2 aims to integrate with existing knowledge bases and SaaS applications to automate various business processes.

Among the features introduced are skills, libraries, and automations. Skills are documented standard operating procedures that outline how the agent harness should perform tasks, and Cohere has made it easier to share these skills for consistency. New automations, which are generic templates customizable via a drag-and-drop graphical interface, allow users to create new skills more efficiently.

Cohere North 2 also introduces shared knowledge and assets called libraries to ensure agents work with the same datasets. Memory is added to maintain context across multiple sessions, similar to macros that work across multiple applications and data sources. The company emphasizes that enterprises can deploy North 2 on-prem, in the cloud, or a hybrid of both, providing flexibility.

Additionally, Cohere North 2 features guardrails to help manage token spend and access controls, with flow control to monitor token usage by departments and access control lists to restrict access to sensitive data. Cohere North 2 represents one of many enterprise-focused agentic AI platforms launched in the past year, with major cloud providers increasingly competing in this space.