Securing the agentic frontier: telemetry, trust and machine-speed defence in the public sector

Executive summary:

The transition from assistive Generative AI (GenAI) chatbots to autonomous, goal-driven Agentic AI represents the most significant shift in enterprise technology since cloud migration. Unlike passive models, agentic systems possess agency: they decompose complex objectives, execute multi-step tool calls, access databases, and make real-time decisions without continuous human intervention.

For critical public infrastructure, policing, and central government, this autonomy unlocks transformative productivity. It also fundamentally rewires the threat landscape across three critical dimensions.

1. Countering autonomous adversaries: defence at machine speed

The Challenge: Cyber adversaries are rapidly weaponising agentic frameworks to conduct autonomous reconnaissance, discover unknown zero-days, coordinate polymorphic attacks, and chain exploits across identity boundaries at speeds impossible for manual human analysis to match.

The Splunk/Cisco Response: Defending against autonomous threats requires continuous, end-to-end data fidelity. Siloed alerts must be replaced by unified streaming telemetry connecting endpoint, network, and cloud environments (Cisco Security + Splunk Enterprise Security), establishing dynamic behavioural baselines that detect and isolate malicious agentic choreography before impact.

2. The internal frontier: governing and securing deployed agents

The Challenge: As organisations deploy AI agents to handle sensitive workflows, they introduce non-human identities with broad privileges. Risks include prompt injection, indirect context contamination, unauthorised lateral tool execution, and unmonitored data exfiltration.

The Splunk/Cisco Response: You cannot secure what you cannot observe. AI agents must be integrated into standard enterprise observability and security architectures. Splunk enables end-to-end logging and telemetry of model inputs, API transactions, decision trees, and downstream actions. This enforces Zero Trust least-privilege principles and verifiable audit trails for AI workloads.

3. Transforming the SOC: the AI-empowered cyber defender

The Challenge: Public sector and commercial Security Operations Centres (SOCs) face chronic skills shortages and alert fatigue, creating asymmetric advantage for attackers.

The Splunk/Cisco Response: By deploying agentic AI defensively using Splunk AI Assistant and Agentic Security Orchestration, Automation and Response (SOAR) (Automation Builder), defenders regain the upper hand. Autonomous agents handle rapid triage, alert correlation, and contextual enrichment in seconds, freeing cyber analysts to focus on complex threat hunting and high-level decision making with human-in-the-loop oversight.

Summary and techUK call to action:

Achieving true digital resilience in the AI era requires a holistic strategy: observing every interaction, governing every autonomous identity, and augmenting human expertise with machine-speed defence.