9robots Scout LLM
Stage: In developmentCybersecurity model retrained weekly — latest threat intel baked into the weights, no inference-time lookups.
What it is
Cybersecurity model retrained weekly — CVEs, attack patterns, and vulnerability disclosures continuously fine-tuned into the weights. No inference-time internet lookup.
Why it matters
Scanning hundreds of thousands of data flows can't afford an internet roundtrip per request. Scout carries the latest threat intel in its weights — zero-latency security analysis at production scale.
How it works
Powers GlassHood infrastructure monitoring with specialised cybersecurity awareness. Also available standalone via ColdVault API for AI security scanning of configs, logs, and network traffic.
Where we are today
In development. Each version is date-stamped (e.g. Scout 2026-05-24). Public API has a one-month delay; direct clients receive the latest weekly build.
What it is
A cybersecurity model retrained weekly on the latest threats. CVEs, attack patterns, and vulnerability disclosures are continuously ingested and fine-tuned directly into the model weights. At inference time there is no internet search — the knowledge is in the model itself.
Why it matters
When scanning hundreds of thousands of data flows through integration pipelines, you cannot afford an internet lookup per request. The cost is wrong twice — the latency tax on every request, and the data-egress risk every time a system asks an external service for help. Scout carries the latest threat intelligence in its weights, enabling zero-latency security analysis at production scale without the egress.
How it works
Powers GlassHood infrastructure monitoring with specialised cybersecurity awareness. Also available standalone via the ColdVault API for organisations that want AI-powered security scanning of configurations, logs, and network traffic. Annex 22 ready.
Where we are today
In development. Freshness guarantee: each version is date-stamped (e.g. 9robots Scout LLM 2026-05-24). Public API availability includes a one-month delay so the wider ecosystem can validate before release; direct clients receive the latest weekly build.