Harvey Completes PROTECTED-Level IRAP Assessment for Australian Agencies
Harvey completed a PROTECTED-level IRAP assessment on November 16, 2026, plus NSW supplier approval and a Victorian marketplace listing — evidence for agency authorisation, not certification.

Why it matters
- Harvey announced on November 16, 2026 an IRAP assessment at the PROTECTED level by Schellman Compliance, LLC.
- A July 30, 2026 Trust Center update had already recorded the assessment without the PROTECTED-level detail.
- Harvey is an approved supplier under the NSW ICT Services Scheme (SCM0020) and listed on Victoria's Digital Marketplace for eServices.
- Australian processing in Microsoft Azure is available, and Harvey does not train on customer data by default.
- IRAP assessment is evidence for agency authorisation, not government certification or endorsement.
Harvey announced on November 16, 2026 that it had completed an independent assessment under Australia's Infosec Registered Assessors Program (IRAP) at the PROTECTED level, alongside supplier approval in New South Wales and a marketplace listing in Victoria. The move gives Australian government legal teams a clearer path from security review to procurement — but it stops short of government certification, and the final authorisation decision still sits with each agency.
The assessment, conducted by Schellman Compliance, LLC, tested the effectiveness of applicable controls in the Australian Government Information Security Manual (ISM) at the PROTECTED level. Harvey says the resulting evidence is intended to inform an organisation's own security risk review and its decision to authorise the system. That distinction matters: an IRAP assessment is evidence for an authorisation decision, not government approval of Harvey's platform.
What does the IRAP assessment actually establish?
The short answer: evidence, not endorsement. The IRAP entry in Harvey's Trust Center names Schellman Compliance, LLC as the assessor and expressly distinguishes an IRAP assessment from a certification.
Official IRAP guidance from the Australian Signals Directorate sets the boundaries. Assessors independently evaluate security posture, identify risks and suggest mitigations. They do not grant government accreditation or endorsement, and they do not register systems. An assessment may also leave some ISM controls outside its scope, and completing one is not itself a finding that the system meets the controls tested.
For an agency, this means the assessment is one input into its own risk decision — not a stamp of approval from Canberra.
The disclosure itself is not entirely new. Harvey's July 30, 2026 Trust Center update had already recorded the assessment's completion against the ISM, without specifying the PROTECTED level. The November 16 announcement adds that detail along with the procurement information.
Takeaway: the assessment gives agencies documented evidence to review, but the authorisation decision remains theirs.
Why does this matter for the Australian market?
Government legal work — research, document review and drafting for complex matters and public services — is a high-stakes market for legal AI vendors. PROTECTED is the classification level at which much of that sensitive government information is handled, so an assessment at that level is a practical prerequisite for serious consideration by agencies.
Harvey reports existing work with multiple public-sector organisations, public universities and government-owned enterprises across Asia-Pacific. The company frames its regional strategy around local fit rather than a one-size-fits-all product.
Ashleigh Whittaker, Harvey's Director of APAC Strategy and Country Manager for Australia and New Zealand, said the company's approach involves "understanding the priorities for each organisation, building a platform that complies with local requirements and working alongside each legal department as they adopt AI."
Takeaway: the announcement signals Harvey is building deliberately toward deeper government adoption in Australia and the wider region.
Which procurement routes are now open?
Two separate channels now exist for eligible buyers, and they solve different parts of the adoption puzzle.
- NSW: Harvey is approved as a supplier under the NSW Government ICT Services Scheme (SCM0020).
- Victoria: Harvey is listed on the Victorian Government's Digital Marketplace for eServices.
Both are presented as established purchasing channels. Paired with the IRAP assessment, they address separate steps in government adoption: reviewing security evidence, and buying through an existing procurement mechanism. Neither status replaces the buying organisation's own authorisation decision.
Takeaway: agencies can now review the security evidence and purchase through standard channels — but they still have to sign off themselves.
How does Harvey handle agency data?
For government buyers, data handling is where the practical questions live. Harvey's security documentation describes a cloud environment hosted in Microsoft Azure, with Australian processing available for customers with localisation requirements or preferences. The selected processing location also applies to subprocessors.
Customers have direct control over several dimensions of data handling:
- Choosing which documents to upload
- Setting retention policies
- Deleting data
- Keeping information in-region
Other listed regional processing options cover the EU or Switzerland and the US.
On the model side, Harvey says data is encrypted at rest and in transit, each customer's data is logically separated, and access follows least-privilege controls. By default, Harvey does not train on customer data. External model providers are contractually barred from training on that data and must operate with zero data retention, while security obligations are passed to subprocessors.
Takeaway: Australian processing, customer-controlled retention and contractual bars on model training form the core of the data story agencies will scrutinise.
What ongoing security practices back the assessment?
An assessment is a snapshot, so the operational picture matters too. Harvey reports:
- Automated vulnerability scans
- Annual third-party penetration tests
- Continuous security monitoring
Its Trust Center offers certifications, security practices and control documentation to current and prospective customers conducting due diligence. That gives agencies material beyond the announcement with which to examine the service — a point worth stressing, since the assessment alone answers only part of the security question.
Takeaway: agencies can dig into the Trust Center documentation rather than relying on the press release.
What happens next?
For Australian government legal teams, the practical change is a clearer route from security review to procurement. The assessment evidence, the NSW supplier approval and the Victorian marketplace listing together cover both halves of the adoption process.
The remaining decision is specific to each organisation: whether the assessed controls, processing locations, subprocessor arrangements and retention settings fit the legal matters and information it intends to place in Harvey. That judgement — not the IRAP assessment itself — is what will determine how far the platform reaches into Australian government legal work.
Source: unite.ai
More from Sophie Lindqvist
Show full bio
Staff writer covering regulation, policy and their impact on artificial intelligence and machine learning at AI In Context.
749 articles
Related articles
- Harvey clears IRAP PROTECTED-level assessment for Australian government work
- Semperis launches Lightning platform to secure human and AI identities
- AI Security in 2027: Agent Permissions Will Define the Blast Radius
- Only 15% of APAC boards feel equipped to oversee major risks
- Australia 'sleepwalking into another robodebt' over government AI use