AI Agent Tools promise to suck up some of the dragery from their daily workflows, but most organizations are still hesitant to adopt them and have immediate concerns: data security. Large companies with trade secrets, highly regulated industry businesses, and government agencies have thought more than once about drawing AI tools out of concern.
Canadian AI Firm Cohere aims to alleviate these concerns with a new AI agent platform called North. This promises to enable private deployments so that businesses and governments can keep their and their customers’ data safe behind their own firewalls.
“LLMS is just as good as the data you can access,” Cohere co-founder Nick Frosst said at a demonstration in North. “If you want LLM to be as convenient as possible, you need to access that useful data, meaning it needs to be deployed. [the customer’s] environment. “
Instead of using enterprise cloud platforms like Azure or AWS, Cohere says it can be installed north on your organization’s private infrastructure to avoid viewing or interacting with customer data. The North can run in your organization’s on-premises infrastructure, hybrid cloud, VPCS, or air-gap environments, Frosst said.
“We can literally deploy it to a GPU in our closet.
Cohere claims the North also includes security protocols such as granular access control, agent autonomy policies, continuous red teaming and third-party security testing. It also meets international compliance standards such as GDPR, SOC-2, and ISO 27001.
More than private development
Cohere, which has raised $970 million so far, has recently been piloting north along with some customers, including RBC, Dell, LG and Ensemble Health Partners, at a valuation of $5.5 billion, and said it is already piloting Palantir north, as TechCrunch reported last year.
TechCrunch Events
San Francisco
|
October 27th-29th, 2025
North is quickly reflecting many AI agent platforms. Its main features are chat and search, allowing users to respond to customer support inquiries. Summary transcript meetings, create marketing copies, and access information from both internal resources and the web. Frosst added that all answers include a quote and a “inference” chain, allowing employees to audit and verify the output.
The chat and search features include existing Cohere technologies such as commands (family of generated AI models) and compasses (multimodal search technology stack). Frosst said the North carries a variant of the command model that is trained for enterprise inference.
“It goes beyond just Q&A and you’ll be working for you. So [North] There are a lot of asset creations. You can create tables, documents, and slideshows. You can do a lot of market research,” Frosst said.
It is worth noting that in May, Cohere acquired Ottogrid, a Vancouver-based platform that develops enterprise tools to automate high-level market research.
Like other AI agent platforms, North can connect to existing workplace tools such as Gmail, Slack, Salesforce, Outlook, Linear, and integrate with any Model Context Protocol (MCP) server to access industry-specific or in-house applications.
“When you gain confidence by chatting with a model, there’s something like a smooth transition that happens between using this, from using it as an automation,” Frosst said.
Fix: Previous versions of this article misunderstood the title of Frossst. I regret the error.
Source link
