AI Agent Security: Obsidian Security Raises $85m for Enterprise Governance

What happened

Obsidian Security has closed an $85 million funding round to expand its platform for detecting and containing threats posed by AI agents operating inside enterprise environments. The raise signals growing institutional concern that the rapid deployment of autonomous AI systems — tools increasingly used to handle customer interactions, process data and execute workflows on behalf of organisations — is outpacing the security controls designed to govern them.

Obsidian's platform is built to monitor the behaviour of non-human identities: AI agents, service accounts and automated pipelines that now sit alongside human employees inside corporate SaaS and cloud environments. The company's thesis is that as organisations delegate more consequential tasks to AI agents, those agents become high-value targets — and potential sources of accidental or malicious harm — if left unmonitored.

Why it matters

For customer experience and service design practitioners, this funding round is a signal worth taking seriously. AI agents are no longer a future consideration; they are already being deployed to handle customer queries, personalise offers, process claims and manage service queues. Each of those agents operates with delegated authority — access to customer data, transaction systems and communication channels. A rogue or compromised agent does not just create a security incident; it creates a trust incident, one that customers experience directly and personally.

From a behavioural economics perspective, the asymmetry here is stark. Customers extend trust to a brand, not to the underlying infrastructure. When an AI agent behaves unexpectedly — whether through a security breach, a misconfiguration or an emergent failure — the attribution lands on the organisation, not the technology vendor. The reputational and relational cost of a single high-profile AI agent failure can dwarf the operational savings that motivated the deployment in the first place. Governance of AI agents is therefore not a back-office IT matter; it is a front-line CX risk.

By the numbers

• $85 million — the size of Obsidian Security's latest funding round, as reported by FinTech Global.

The Renascence take

Most CX leaders reading about this round will file it under "cybersecurity" and move on. That is the wrong instinct. The more interesting question is not whether AI agents can be attacked from outside, but whether organisations have any real visibility into what their own agents are doing to customers right now.

The conversation about AI in customer experience has been dominated by capability — what agents can do — while almost entirely ignoring accountability — what they are actually doing, to whom, and with what effect. Obsidian's raise reflects a market waking up to the governance gap, but the tools being funded are security tools, not experience tools. Customer-obsessed operators should be asking a harder question: before we worry about external threats to our AI agents, do we even have the observability infrastructure to know when an agent has delivered a bad, biased or damaging experience? Trust is built in interactions, not in infrastructure audits — and right now, most organisations cannot answer that question with any confidence.