See every AI interaction. Investigate 30× faster.
Anjung Sentinel is the local AI SecOps console for LogRhythm SIEM. It monitors enterprise AI use and Claude Enterprise agents, and lets EvolveX AI investigate with models that run on your own hardware.
One console for AI-era security operations
Built on the LogRhythm data you already collect. Ten modules cover shadow AI, rogue agents, investigation, detection, local inference and proof for auditors.
Enterprise AI activity
Every prompt to ChatGPT, Copilot, DeepSeek, browser extensions and internal LLMs, parsed from the LogRhythm sources you already collect, scored for risk.
Claude Enterprise agents
Each agent gets a 30-day behavioural baseline. Drift is scored every 15 minutes, and a rogue agent can be suspended from the console.
EvolveX AI investigator
A persistent investigator that gathers context, runs follow-up searches and posts a cited triage summary within about 10 minutes.
Related Cases
Cases sharing a key, host, IP or agent are linked with a confidence score and a plain-language reason, then merged in one click.
Agentic SOC plugin
Triage alerts, prioritise your shift and investigate cases in natural language from Claude Code or OpenAI Codex.
AI detection rules
Seven rules for AI-era threats mapped to MITRE ATLAS and ATT&CK, plus plain-language rule drafting with backtest and two-person approval.
Local models and guardrails
Run Llama, Qwen or Mistral on your own GPUs. Secrets are masked, context follows analyst permissions and actions need a human.
Outcomes and Executive Digest
A 250-word digest for leadership and an Outcomes Navigator that ties SOC work to the business priorities that matter.
Tamper-evident audit trail
Every model request and AI-driven action is SHA-256 hash-chained, forwarded to LogRhythm and exportable for SOC 2 and ISO 27001.
How it works
From raw log record to a contained incident, without a single prompt leaving your network.
Collect
Proxy, firewall, endpoint, CASB, DLP, SaaS and AI gateway logs flow through LogRhythm as they do today.
Classify and score
AI interactions are recognised, secrets are masked at ingest, and each event gets a High, Medium or Low risk.
Investigate
EvolveX AI opens the case, gathers context, links related cases and answers questions with citations.
Prove
Every request and action is hash-chained in the audit trail and summarised for leadership in the digest.
Faster investigation with EvolveX AI
EvolveX AI works across the New-Scale Security Operations Platform as a persistent investigator: it gathers context when a case opens and keeps searching as the incident unfolds.
- Every answer cites its log sources and record counts
- Says "I can't tell from the evidence" instead of guessing
- Suggests containment; a person always approves it
Average case triage time
As measured by our own security operations team.
Security you can prove
Guardrails that are locked on, not promised.
Block all external model calls
Network policy denies egress from inference and app pods. An egress counter watches firewall logs every 10 seconds.
Mask secrets and personal data
Keys, passwords, tokens, national IDs and card numbers are replaced with reversible tokens before inference.
Limit context to permissions
The model only sees log sources the analyst can already search in LogRhythm.
Require human approval
Any model output that maps to an action becomes a proposal. Only a person can run it.
Bring AI under control in your SOC
Request access to the Anjung Sentinel Console. An administrator reviews every request.