A law practice is a data business that happens to argue in court. Every matter file holds the raw material of people’s lives — contracts, financial records, affidavits, correspondence that was never meant for other eyes — and the profession’s confidentiality duty makes protection non-optional. What has been missing is a method. DataSecOps supplies one: the approach defined on 4 August 2021 by Ben Herzberg during his Satori Cyber chief-scientist years, embedding security into the everyday flow of changing data and users; a vendor’s coinage rather than a statutory term, but the discipline behind it — long known as data security governance — maps onto a law office unusually well, because lawyers already understand chain of custody, access on need, and documented authority. The seven steps below convert that kinship into an implementable plan, each with the evidence it should produce.

Step 1: Inventory the matter files

Begin where any audit begins: with the record. List every location client material lives — the practice management system, the shared drive, individual laptops, the backup drive in the storeroom, the scanner’s memory, the cloud folder a junior set up in 2023. For each location, note who can reach it and how. The product of this step is a single document, dated and signed, that the practice can produce to any client or court asking where their information resides. Practices that skip this step later discover shadow copies during disputes, at the worst possible moment.

Step 2: Classify by sensitivity, in writing

Adopt a written classification scheme with three tiers at minimum: public (filed judgments, published work), internal (drafts, research), and confidential (client identities, financial records, settlement positions). Attach the classification to the file, not to the folder — a matter containing settlement authority is confidential even when stored among routine papers. Automated classifiers can propose labels where documents are digital, and their limits are documented: they never achieve full accuracy, so a named person confirms what is flagged IDENTIFIER or SENSITIVE. Where drafts are shared externally, note the boundary drawn by Europe’s framework: Article 4(5) defines pseudonymised data as information that can no longer be attributed to a person without separate kept-separately information — useful when arguing about what a redacted draft actually reveals.

Step 3: Write the DataSecOps rules as versioned documents

The engineering habit worth copying wholesale is that rules live as artefacts with authors and dates, not as habits in someone’s head. A practice should hold a written access matrix: which role opens which tier of file, approved by the managing partner, revised through a documented change. When a dispute arises about who was permitted to see a settlement note in March, the practice answers with the matrix and its revision history rather than with recollection. This is the same discipline software teams call policy-as-code, implemented in whatever tooling the office actually uses — the medium matters less than the existence of an auditable trail.

Step 4: Enforce at the smallest unit that works

Perimeter thinking fails here as everywhere: a password on the front door of the shared drive says nothing about which associate may open which file. Enforcement belongs at the finest practical unit — the individual matter file, ideally the document. The associate on the tenancy matter sees tenancy files; the tax specialist sees tax files; the intern sees whatever the responsible partner explicitly enumerated, and nothing else. Technology exists at every price point; the non-negotiable part is that the restriction applies when the file is opened, not merely when it is stored.

Step 5: Control the outflows

Case material leaves constantly: briefs to counsel, documents to opposing counsel, updates to clients — much of it by email and messaging apps. An outflow policy names the approved channels, requires the confidential tier to travel encrypted or through a portal, and logs each external send. Bulk export deserves special suspicion: a departing associate downloading the entire matters database is an exit event the practice should see in real time. Most breaches in professional services are exits, not break-ins; the evidence-led practice treats every outflow as recorded fact, not as assumed propriety.

Step 6: Keep and read the access log

Every system in the practice records who opened what. Reading that record weekly converts it from forensic debris into an early-warning instrument: a clerk browsing files outside their matters, an account active at midnight, a pattern of failed attempts. Two honest limitations deserve note, because analytical rigour includes knowing what the record cannot say. Database-level audit tools such as pgAudit cannot reliably capture the actions of superuser accounts — the most privileged actors are partly outside the log — and histories carry latency, so the freshest hours may be incomplete. Compensate by narrowing administrative accounts to the smallest viable list and reviewing their console activity through the provider’s own records.

Step 7: Run the DataSecOps review that revokes

Access rots: associates leave, matters close, retainers end, and permissions linger past their purpose. The review is a calendar event with two questions per entry — does this person still need this file, and has it been used — followed by deletions. A review that revokes nothing is a ceremony, not a control; the professional standard is at least one documented revocation per cycle, or a written explanation why none was warranted. Grant anything sensitive with an end date attached, so the expiry does the quiet work between reviews.

The framework context, briefly

Practices serving European clients will recognise the duties above in GDPR Article 32 (security of processing) and Article 25 (data protection by design) — this plan is essentially those articles rendered as office procedure. The full DataSecOps method, including the maturity ladder and the documented limits of each engine’s controls, is maintained in the Ops Atlas DataSecOps entry. Adjacent reading from our own pages: the infrastructure layer underneath is covered in cloud security and DevSecOps for modern applications; the data-team roles behind step two are profiled in professional DataOps training and certifications; and the transformation context for firms modernising end to end is drawn in DevOps in enterprise digital transformation.

  • Inventory first, in writing, dated and signed — shadow copies surface during disputes otherwise.
  • Three tiers, attached to files not folders, humans confirming machine labels.
  • Rules as artefacts with authors and dates; recollection is not a control.
  • Enforce at the matter file, not the front door.
  • Log every outflow; bulk exports are exit events, watched in real time.
  • Weekly log reading, narrowed admin accounts, quarterly reviews that revoke.

Leave a Reply

Your email address will not be published. Required fields are marked *

Facebook Twitter Instagram Linkedin Youtube