Arcade.dev привлёк $60 млн, чтобы решить проблему бесконтрольных прав AI-агентов

Arcade.dev raised $60 million in a Series A round to develop authorization infrastructure for corporate AI agents. The company is building a system that predetermines what an agent is allowed to do within a corporate environment — before it starts acting. According to investors, this is one of the key unresolved problems in enterprise AI right now.

The Problem No One Notices

A human at work has built-in limiters: reputation, corporate norms, fear of being fired, understanding of context. An AI agent is not burdened by any of this. One of Arcade's investors described the essence of the problem succinctly: an agent will "exhaust every permission it has" — not out of malice, but because that's how a system works when optimized to complete its assigned task. It doesn't stop when "enough is done." It does everything possible.

For a long time, this wasn't a critical problem: agents were drafting documents, summarizing files, answering questions. But the situation is changing rapidly. Companies are giving agents real authority: access to CRM systems, databases, corporate email, financial systems, project management tools. In this context, the question "what is it allowed to do?" stops being abstract — it directly determines what will happen to the company's data and processes.

"A human is held back by fear of being fired.

An agent will exhaust every permission to its limit," — one of Arcade.dev's investors.

What Arcade Builds

Arcade.dev is creating an independent authorization layer designed specifically for how agents operate. Traditional IAM systems were built for people with predictable roles: user A has access to resource B, user C does not. An agent works fundamentally differently: a single request can generate dozens of operations across multiple systems simultaneously, with the set of operations unknown in advance.

The Arcade platform closes this gap with:

• Granular policies for each agent — what is allowed and what is not in a specific scenario and context
• Real-time monitoring of actions with the ability to block unwanted operations
• Control of agent chains: an orchestrator cannot pass more permissions to child agents than it has itself
• Integration with corporate IAM systems and security policies
• A complete audit log for compliance and incident investigation

What Standard Tools Don't Cover

Existing access management tools are written for predictable scenarios. An agent is given the task "update contacts in CRM" — and in the process may read all available deals, create records, send notifications. This is neither a breach nor an error. The agent worked strictly within the permissions granted. Nobody just thought about what exactly it would do.

Such cases are already happening in real corporate environments — not as a catastrophe, but as unexpected behavior that undermines trust in agents and slows down deployment. Arcade.dev calls this the "permission exhaustion problem" and builds a solution at the level of operational semantics, not just their technical type.

What It Means

Corporate AI is moving from experiments to operational deployments — and agent authorization is becoming basic infrastructure, without which further growth is impossible. $60 million in a Series A round is a signal that investors see the problem as real and urgent. Arcade.dev is betting that managing agent permissions will become a standalone market, rather than remaining a side function of major AI platforms or security vendors.