ESET unveils AI agent and conversation security tools at Black Hat
ESET used the Black Hat 2026 conference in Las Vegas to announce three new security capabilities aimed at the risks introduced by autonomous AI agents and generative AI tools in the enterprise. The features, due to become available on 30 September across the ESET PROTECT portfolio, cover AI supply-chain inspection, agent behavioural monitoring and generative AI conversation safety.
The company says the additions are embedded directly into its existing platform rather than offered as separate paid add-ons, which it frames as a differentiator for channel partners selling into cost-conscious IT teams. ESET did not disclose pricing changes, updated annual recurring revenue, or named customer deployments in its release.
The three new capabilities
ESET AI Agent Security is designed to inspect components across the AI supply chain as autonomous agents access files, call external services, or interact with code repositories. The intent is to catch malicious or compromised elements before they reach the customer environment.
AI Behavioral Monitoring sits alongside Agent Security but focuses on runtime actions: it is intended to detect and block agents that attempt to access inappropriate resources or behave outside their defined scope. The third capability, AI Conversation Security, targets generative AI tool usage by inspecting uploaded files and metadata for sensitive content and flagging AI-generated responses linked to phishing or malicious sites.
Kamil Pšenák, Senior AI Product Manager at ESET, framed the rationale plainly: "AI is a new class of actor inside the company, reading, writing, making decisions and executing. As such, it deserves the same security attention as users and endpoints, because it behaves like both at once." ESET says it owns the underlying models and intelligence rather than licensing third-party components.
Market context
The launch arrives as shadow AI, where employees adopt AI tools without IT oversight, has become one of the faster-growing enterprise security concerns. A number of established endpoint and cloud security vendors have moved to address agentic AI risk in the past twelve months, including CrowdStrike, SentinelOne and Palo Alto Networks, each extending their existing platforms to cover AI workload behaviour. Pure-play AI security startups have also attracted significant venture funding in the same period, making this one of the more crowded emerging segments in cybersecurity.
ESET's positioning is that two decades of machine-learning research gives it an advantage in threat modelling for AI-specific attack vectors. That claim is plausible on the prevention side, where the company has a credible long-run record, but the behavioural monitoring and conversation inspection categories are relatively new to any vendor's portfolio, and buyers will want benchmark evidence before drawing conclusions.
Regulatory read-across
The EU AI Act is directly relevant here. Its obligations for high-risk AI systems and general-purpose AI models include requirements for human oversight, logging of autonomous agent actions, and transparency about AI-generated outputs. Enterprises deploying AI agents in regulated sectors, including financial services, healthcare and critical infrastructure, face compliance deadlines that will accelerate demand for the kind of runtime monitoring ESET is now offering. The UK's emerging AI governance framework and the US NIST AI Risk Management Framework point in similar directions. Vendors that can map their controls to these frameworks are likely to win procurement conversations faster in the enterprise segment.
ESET is also showcasing a joint capability with Intel at the event, using processor-level execution telemetry from Intel DTECT alongside its own AI models to improve visibility in cloud workloads. ESET PROTECT's next significant milestones will be the 30 September general availability date and any named customer wins that emerge from Black Hat conversations.