KaribuTech AI / Expertise
AI Cybersecurity
Protect AI systems and use AI responsibly in security operations, from threat modelling to incident response.
LLM Security & Prompt Injection
Treat prompts, uploaded documents and retrieved pages as untrusted data. Keep secrets outside model context, validate outputs and enforce permissions in application code rather than relying on model instructions.
Agent & Connector Security
Use narrow tool permissions, restricted destinations and explicit approval for consequential actions. Protect against excessive agency, data exfiltration and malicious tool responses.
AI Threat Detection & SOC
Assist analysts with event correlation, anomaly triage and evidence summaries. Validate alert precision and false positives on representative logs; do not let unverified generated recommendations trigger destructive actions.
Model & Data Supply Chain
Track dataset provenance, model sources and dependencies. Review artefacts for tampering, poisoning and unsafe loading; isolate evaluation environments and retain signed or verified deployment records where supported.
AI Red Teaming & Incident Response
Run authorised tests for prompt injection, leakage and permission bypass. Record reproducible findings, severity and remediation ownership, and rehearse isolation, credential rotation and recovery.
Insights & research
AI Cybersecurity articles
| Article | |
|---|---|
| Draft | AI cybersecurity manual: prompt injection and secure tool access Threat-model enterprise assistants and prevent untrusted content from becoming privileged instructions. AI Cybersecurity |
| Draft | AI-assisted security operations and model supply-chain controls Connect analyst assistance with evidence quality, model provenance and safe incident handling. AI Cybersecurity |