Finds personal and confidential data in Malay, English and mixed text.
- ✓Redact personal data before an LLM call
- ✓Scan documents and chat logs
- ✓Feed Guardian PII guardrails
TypeToken classifierSize0.3BRuns onCPU only · 2 vCPUDeployCloud API · On-prem
A fast CPU filter for toxic, profane, sexual and jailbreak content in Malay, English and mixed text.
- ✓Cut judge cost by filtering first
- ✓Screen every prompt on CPU
- ✓Malay, English and mixed
TypeEmbedding classifierSizeEmbedding + headsRuns onCPU only · 4 GB RAMDeployCloud API · On-prem
Detects prompt-injection attempts in user input and retrieved documents.
- ✓Protect RAG pipelines
- ✓Guard agent tool inputs
- ✓Block hidden instructions
TypeEncoder classifierSize0.1BRuns onCPU only · 1 vCPUDeployNot yet released
On the roadmap. Tell us if you need it.
Judges text against a written control. Give it the policy in plain language and it returns a verdict.
- ✓Singapore and Malaysia PDPA controls
- ✓PCI DSS sensitive-data checks
- ✓GDPR and Malaysian food-claim controls
- ✓The “own rule” in Protections
TypeSmall generative judgeSize4BRuns onGPU 8 GB (4-bit) or CPU quantisedDeployCloud API · On-prem
Gives a verdict with a written rationale, for the audit trail. Reasons in Malay and English.
- ✓Escalated and borderline cases
- ✓Audit-ready explanations
- ✓Reviewer assistance
TypeReasoning LLMSize8BRuns onGPU 16 GBDeployNot yet released
On the roadmap. Tell us if you need it.
Maps your controls to the clauses of PDPA, RMiT and PCI-DSS.
- ✓Build a control-to-clause matrix
- ✓Show coverage gaps
- ✓Answer an auditor's "where is this covered?"
TypeRetrieval modelSizeEmbedding + rerankerRuns onCPU only · 4 GB RAMDeployNot yet released
On the roadmap. Tell us if you need it.
Checks each sentence of an answer against the source it cites.
- ✓Grounded-answer checks for any RAG app
- ✓Catch unsupported claims
- ✓Regulatory content review
TypeFact-check classifierSize0.1BRuns onCPU only · 2 vCPUDeployNot yet released
On the roadmap. Tell us if you need it.
Detects leaked recipes, formulas and secrets by similarity, without sending them anywhere.
- ✓Trade-secret leak detection
- ✓Fingerprint your own documents
- ✓Works fully offline
TypeN-gram and embedding matcherSizeMatcherRuns onCPU only · 2 GB RAMDeployNot yet released
On the roadmap. Tell us if you need it.
Drafts grounded articles and finding write-ups that stay close to their sources.
- ✓Regulatory news drafts
- ✓Scout and SOC finding write-ups
- ✓Always cites its sources
TypeGenerative LLMSize8B–14BRuns onGPU 16 GBDeployNot yet released
On the roadmap. Tell us if you need it.
Triages SOC alerts and writes the summary an analyst would write.
- ✓Alert triage
- ✓Incident summaries
- ✓Runs inside the SOC network
TypeReasoning LLMSize8BRuns onGPU 16 GBDeployNot yet released
On the roadmap. Tell us if you need it.
Turns a plain-language question into a query over your logs and governance data.
- ✓Ask your audit trail
- ✓ClickHouse and Postgres
- ✓Read-only by design
TypeCode-capable LLMSize7BRuns onGPU 8 GB (4-bit)DeployNot yet released
On the roadmap. Tell us if you need it.
An air-gapped assistant that explains incidents and findings to your team.
- ✓Ask Obi, offline
- ✓Incident narration
- ✓No data leaves your network
TypeGenerative LLMSize8B–20BRuns onGPU 16 GBDeployNot yet released
On the roadmap. Tell us if you need it.
A model tuned on your own data and policies, delivered with a benchmark on your cases.
- ✓Your terminology and document types
- ✓Retrained on a schedule
- ✓Benchmark report included
TypePer-customer adapterSizeLoRA adapterRuns onSame as the base modelDeployNot yet released
On the roadmap. Tell us if you need it.
Hardware figures are typical sizing. We confirm them for your volume during the pilot.