Building Legal AI for Public Service Across Borders

By Joshua McKibben, Head of Trust, Harvey.

Harvey is investing in public sector legal teams across jurisdictions. Our recent FedRAMP Class D (High) certification in the United States and completion of an independent IRAP assessment at the PROTECTED level in Australia mark important steps in that work. These milestones reflect our commitment to giving public sector counsel more capacity to serve their communities and providing the security evidence their institutions need to adopt AI with confidence.

Public sector legal work spans national agencies coordinating across departments, state and provincial teams advising across public programs, and city attorneys handling procurement and disputes with small teams. It also supports public universities’ research and education missions and the essential services delivered by publicly owned utilities. Across these institutions, legal teams support consequential decisions, and their ability to adopt AI depends on how well it fits their work, responsibilities, and requirements.

More Capacity for the Work Behind Public Service

A procurement agreement waiting for legal review can hold up a public project. A records request can require an attorney to work through years of correspondence. A policy proposal may depend on interpreting legislation alongside earlier advice scattered across departments.

Before counsel can advise, they must understand the record, identify the relevant requirements, and establish what the organization already knows. That preparation demands time and attention alongside the legal analysis itself.

AI can support the work that comes before a decision. In procurement, it can compare agreements against supplied requirements, surface deviations, and prepare suggested edits for counsel to review. In a public records matter, it can identify potentially responsive material with references to the underlying documents. Attorneys retain responsibility for disclosure, exemptions, and privilege.

For investigations and litigation, AI can extract facts from documents, organize them into chronologies, compare accounts, and surface gaps for further examination. A clearer starting point gives counsel more time to evaluate evidence, test an interpretation, and develop advice for the people responsible for acting on it.

Put Shared Legal Experience to Work

A team’s ability to advise also depends on how readily it can draw on work already done. A previous legal opinion may explain why a policy took its current form. An earlier contract may show how an unusual provision was negotiated. Supporting analysis may inform a similar question in another department.

Making relevant prior work searchable, with supporting sources, can give attorneys a stronger foundation for current advice. They can examine the earlier reasoning, check whether the law or circumstances have changed, and decide what applies to the question in front of them.

This can support continuity and consistency across a team. It can also give colleagues a way to learn from one another’s work as responsibilities change. The value comes from making experience available for scrutiny and reuse, with permissions that reflect the sensitivity of the material.

Trust Must Fit the Jurisdiction

These opportunities exist across borders, but the conditions for adoption differ. An Australian government department and a US agency may face similar legal workloads while operating under different security, data-handling, and procurement requirements.

Public sector organizations need evidence about how information is protected, where it is processed, who can access it, and how the service fits their obligations. That evidence must be relevant to the organization evaluating it.

Australia’s Information Security Registered Assessors Program, or IRAP, provides one example. An independent assessment examines applicable security controls and provides evidence an organization can use in its own risk assessment and system authorization decision. In the United States, FedRAMP addresses federal cloud security requirements, while TX-RAMP addresses requirements for cloud services used by Texas state agencies and public institutions of higher education.

Security evaluation also needs to connect to daily legal work. Attorneys need sources they can check, administrators need controls over access and information sharing, and teams need clear expectations for review and accountability. Together, these practices support responsible AI use within the institution’s legal work.

A Long-Term Commitment Across Markets

Harvey’s IRAP assessment examined applicable Australian Government Information Security Manual controls at the PROTECTED level. It gives public sector organizations independent evidence to inform their evaluation of Harvey. We have also been approved as a supplier under the NSW Government ICT Services Scheme, providing an established procurement route for eligible buyers.

These steps sit alongside FedRAMP Class D certification in partnership with Knox Systems and TX-RAMP Level 2 certification in the United States. They reflect investment in the requirements of different markets and institutions.

As Head of Trust, I see that work as an ongoing responsibility. Completing an assessment gives us a foundation for conversations with an organization’s security team, and I expect those conversations to continue as its needs and our technology evolve. The teams entrusting us with their information should be able to see how we maintain our controls and respond to new risks over time. That means making security part of our products and operations, providing clear documentation, and being accountable for the answers we give.

Public sector legal teams should be able to expect a long-term partner that understands their responsibilities and continues to earn their trust. That is the commitment we are making across the markets we serve.

Learn more about Harvey for Public Sector, or visit our Trust Center to review our security documentation.

–

[ This is a sponsored thought leadership article by Harvey for Artificial Lawyer. ]


Discover more from Artificial Lawyer

Subscribe to get the latest posts sent to your email.