Proofpoint Opens Hyderabad AI Security Engineering Center

In an interview with Computer Weekly, Proofpoint CEO Sumit Dhawan described AI systems as acting in ways that require security teams to understand the intent behind their actions. He also said agent communication protocols can change rapidly as model providers add capabilities, requiring ongoing development of detection models.

This is a distinct technical problem from conventional code scanning. Agentic systems can invoke tools, access enterprise data, communicate with other services, and execute multi-step workflows. Across the security industry, comparable efforts increasingly combine behavioral telemetry, identity context, data-access controls, and anomaly detection to assess whether an agent's activity is authorized and consistent with an intended task.

Proofpoint already applies intent models to human behavior across email and endpoints, Dhawan told Computer Weekly, but said those models do not directly transfer to machines. The reported Hyderabad work therefore concerns new models for AI-agent activity, rather than a simple extension of existing employee-risk detection.

Proofpoint reported that its India business tripled over the preceding 12 months. The company also reported double-digit growth in new-customer acquisition and a more than 600% increase in protected seats over 18 months. The Economic Times reported that recent customer wins included organizations in banking and financial services, business process outsourcing, defense, and the public sector.

The company's 2026 AI and Human Risk Landscape report found that 87% of organizations globally had deployed AI, while 52% of those organizations were not fully confident that they could detect a compromised AI system. Those figures are company research, not an independent measurement of the broader market, but they identify the operational problem Proofpoint is targeting with the India expansion.

The Economic Times also cited Proofpoint's Voice of the CISO research, which found that 99% of surveyed Indian CISOs reported material loss of sensitive data, compared with 66% globally. The survey result underscores a wider issue for data teams: AI adoption can expand the systems and channels through which sensitive data is created, retrieved, transformed, and shared.

According to The Economic Times, Proofpoint's data-security offering combines Data Security Posture Management (DSPM), Enterprise and Adaptive Email Data Loss Prevention (DLP), and Insider Threat Management (ITM). The company describes these capabilities as tools for discovering and classifying sensitive data, preventing loss across employee and AI-system channels, and detecting risky user behavior.

The Economic Times also reported increased demand for Proofpoint's Active Exploits Protection capability, which is intended to identify vulnerabilities under active exploitation and prioritize remediation according to attacker activity.

For security and ML practitioners, the engineering-centre announcement reflects a broader industry pattern: controls for AI systems are moving beyond model input and output filtering toward monitoring of agent behavior, tool use, identities, and access to sensitive data. Effectiveness in that setting depends on whether telemetry, data classification, and policy enforcement remain usable across rapidly changing models, protocols, and enterprise workflows.