FOCI-ready, PSPF-aligned, built in Sydney
For Australian federal and state government, Tier-1 telcos, and defence-adjacent buyers, the procurement question isn’t “is the AI model good?” — it’s “can we sign off on the supply chain?” This page maps Camlin Speech against the frameworks your procurement team will ask about, before they ask.
What we map against
PSPF Direction 001/2025
Department of Home Affairs ban on DeepSeek across non-corporate Commonwealth entities. The bellwether for "Chinese-origin model = procurement risk".
No Chinese-origin model weights or training data in the production tree. Base TTS shortlist is MetaVoice-1B (UK), Parler-TTS (FR/US), or StyleTTS 2 (US academic). LoRA training data provenance is documented per source.
DTA AI Policy v2.0 (effective 15 Dec 2025)
Mandatory Foreign Ownership, Control or Influence (FOCI) assessments for AI supply chains. Provides model contract clauses for buyers to specify "banned AI systems".
FOCI-ready: Australian-incorporated vendor (Nuamedia Solutions Pty Ltd), Western base-model lineage, no ownership stake or board influence from foreign-FOCI entities. Assessment pack available on request.
Hosting Certification Framework (HCF)
Centralised list of approved AU/foreign companies for OFFICIAL information processing.
Production deployment runs in AWS ap-southeast-2 (Sydney) — HCF-approved hosting region. On-prem option available for tenders requiring air-gapped hosting.
ASD/ACSC guidance
Critical-infrastructure procurement guidance. Telcos in particular track this closely and treat foreign AI similarly to federal gov.
On-prem deployments use only the binaries, model weights and inference code we ship — no inference-time call-out to vendor cloud. The audio path stays inside the customer perimeter.
National AI Plan 2025
"Sovereign capability" called out as a procurement preference.
First-party Australian-English speech engine, fine-tuned on AU-licensed data, recorded in a Sydney studio, delivered by an Australian-incorporated vendor.
Hosting options
- AWS ap-southeast-2 (Sydney). Default. HCF- approved region. Audio data resides and processes in Sydney; nothing routes outside the jurisdiction.
- Customer on-prem. Camlin Speech ships as a set of containers + model weights you run on hardware you already own. No inference-time call-out, no telemetry off the box. Suitable for the most regulated deployments and air-gapped environments.
- Hybrid. ASR on customer hardware, TTS via managed Sydney instance, or any other split that matches your data-handling policy.
Supply-chain transparency
- Western-origin lineage. All model components in the production tree originate from jurisdictions acceptable to AU procurement. No Chinese-origin model weights or training data, regardless of how permissive their licence is, per PSPF / FOCI guidance.
- Australian-trained components. The AU- tuned layer that gets us ahead on Centrelink CRN, rego renewal, NBN FTTP, and AU place names is trained in Australia on AU-licensed data with documented provenance.
- Lineage statement on request. The full supply-chain manifest — pretrained components, their origins, training-data sources and licences, build reproducibility — is included in the FOCI-ready assessment pack we send to procurement teams.
- Sovereignty escalation path. For tenders that require zero foreign-origin pretrained weights at all, we have a documented from-scratch path on AU-only data. Available on request; not the default.
Get the FOCI-ready pack
We’ll respond within one business day with the pack and offer a 30-minute walkthrough with the deployment team.