What Antidote does
Antidote is a three‑part safety platform for AI systems, delivered as a managed cloud service. Use whichever surfaces (or all of them) fit your stack.Data Integrity
Scan training datasets for label errors, outliers, poisoning, bias, leaked
secrets, and prompt‑injection payloads. Triage, heal, and audit the results.
Runtime Security
A drop‑in firewall for LLM traffic. Scan prompts before they reach the model
and responses before they reach the user.
DLP (endpoint)
A signed desktop agent that intercepts LLM traffic at the OS network layer
on every managed laptop. No application code changes.
Pick a starting point
Quickstart
Sign in, scan your first dataset, and make your first Runtime Security call
in about ten minutes.
Data Integrity overview
Core concepts, severity scale, and the map of every feature.
Runtime Security overview
Three ways to integrate the LLM‑traffic firewall, plus authentication.
DLP overview
Endpoint coverage for shadow‑IT AI traffic across the whole workforce.
How the surfaces relate
| Data Integrity | Runtime Security | DLP (endpoint) | |
|---|---|---|---|
| What it inspects | Datasets at rest (images, text, NIfTI, detection) | Live LLM prompts and responses from instrumented apps | Live LLM prompts and responses from any app on a managed device |
| When it runs | On upload, on schedule, on demand | On every request, inline | On every request, on the device |
| Who deploys it | The data / ML team | The application team | IT, via MDM |
| Integration | Web UI, REST API, scheduled scans | Reverse proxy (swap base_url) or REST scan endpoints | Push the signed agent through Jamf, Kandji, Intune, GPO, SCCM |
| Output | Issue lists, healed datasets, compliance reports | allow / redact / block verdicts, events, drift charts | Same verdicts, attributed to a device and a directory user |