Claude vs on-premise AI: which does your compliance posture permit?

Claude is an excellent model, and if your compliance posture permits cloud AI, it may be all you need. Dominion exists for businesses whose obligations mean client data cannot leave the premises at all. This page is for working out which of those two situations is yours.

What Anthropic gets right

  • No training on your data. Anthropic's commercial terms state that Anthropic may not train models on customer content from its services (Anthropic commercial terms).
  • Serious certifications. SOC 2 Type II, ISO 27001 and ISO/IEC 42001 for AI management systems (Anthropic trust centre), with a zero-data-retention addendum available to enterprise customers.
  • A genuine safety culture and enterprise controls that are among the best in the category.
  • Frontier capability that a local model does not match on open-ended reasoning.

The two lines that matter for an Australian practice

First: "not trained on" is not the same as "never leaves your premises". Using Claude means transmitting your material to Anthropic's infrastructure for processing. If your obligations turn on disclosure and third-party access, that transmission is the whole analysis, and it exists regardless of how well Anthropic behaves once the data arrives.

Second, from Anthropic's own commercial terms: for customers outside Europe and the UK, the agreement is governed by the laws of the State of California, with disputes heard in California courts. That is not a criticism of Anthropic; it is simply what signing means. A practice weighing privilege or health information obligations should know which legal system its AI relationship lives under.

Side by side

Claude (enterprise use)Dominion
Trains on your dataNo, per commercial termsImpossible: nothing leaves your building
Where work is processedAnthropic infrastructureOn the unit in your office
Governing law of the relationshipCalifornia, for Australian customersAustralian agreement, Australian hardware
Model capabilityFrontier modelsStrong open models, tuned to your business by ErmosIQ
Cost modelPer seat or usage-based APIPer unit, sized to your practice: our cost page

Dominion is priced per unit, per month on a 12-month term, sized to your practice: unlimited users, no per-user licences, no usage charges. Exact pricing is confirmed in the scoping consultation.

One thing to be clear about

Dominion does not run Claude. The unit runs strong open-weight models that live entirely on your hardware and are tuned to your business over time. If what you need is Claude specifically, you need Anthropic's cloud; if what you need is AI that never leaves the building, that is the trade Dominion makes, openly.

Common questions

Does Claude train on our business data?
No. Anthropic's commercial terms state that Anthropic may not train models on customer content from its services. The question that remains for a regulated business is transmission: using Claude still means sending your material to Anthropic's infrastructure for processing.
Is Claude certified for enterprise security?
Yes. Anthropic holds SOC 2 Type II, ISO 27001 and ISO/IEC 42001, publishes a trust centre, and offers enterprise customers a zero-data-retention addendum. Its enterprise posture is among the strongest in the category.
What jurisdiction governs a Claude agreement for an Australian business?
Under Anthropic's commercial terms, customers outside Europe and the UK contract under the laws of the State of California with disputes heard in California courts.
Can Dominion run Claude on-premises?
No. Anthropic does not offer Claude for on-premises deployment, and Dominion runs open-weight models on your own hardware instead. The choice is between frontier cloud capability and AI that never leaves your building.

See Dominion answering from your own documents.

A demonstration takes 30 minutes, uses no client data, and comes with no obligation.

Book a demo