Skip to content
Back

Where Client Data Is Exposed: A Map of the Risks

Client data can be exposed through devices, networks, applications, stored files, people and AI services. Map six points in a normal accounting workflow and identify simple controls.

Xian Hui

Xian Hui

9 November 2026 · 7 min read

An accountant points to one highlighted point on a wall diagram linking email, cloud storage, laptops, a USB stick, a printer and a phone

In brief

Where can client data be exposed in an accounting firm's workflow?

Client data can be exposed through devices, networks, applications, stored files and backups, human handling, and external AI services. A compromised account, incorrect permission, mistaken recipient or unsuitable sharing decision can put records beyond the firm's intended control. Trace one document through the workflow to identify where safeguards are needed.

What you'll learn

  • Map six exposure points in a client-data workflow
  • Distinguish system access from permission to use or disclose client data
  • Identify simple controls for files, accounts and external services
  • Assess whether an AI tool should receive identifiable client records

Where Client Data Is Exposed: A Map of the Risks

A client sends an accountant a general ledger, bank statements and payroll schedules. The files pass through email, a laptop, a cloud folder and perhaps several applications before the financial statements are prepared.

The files may be needed for a legitimate accounting task. The question is whether each person and system handling them needs the information and has the right safeguards.

What does exposure mean?

Client data is exposed when information becomes accessible, disclosed or usable beyond its intended and authorised scope. That can happen through a deliberate attack or an ordinary mistake.

A risk point is not automatically a breach. For example, using a cloud service is not itself a breach. But an unrestricted sharing link, a compromised account or a mistaken recipient may allow access that was not intended.

The Personal Data Protection Commission (PDPC) Guide to Data Protection Practices for ICT Systems, updated in July 2026, covers information handling, security risk management and safeguarding data through its lifecycle. The NIST Cybersecurity Framework 2.0 small-business guide also encourages firms to identify what information and systems need protection before deciding which safeguards to apply.

Where are the six common exposure points?

Follow a fictional firm's data from receipt to final delivery.

Exposure pointAccounting exampleFirst question to ask
1. DeviceA laptop downloads client payroll dataIs the device secured, updated and access-controlled?
2. NetworkA remote worker connects to shared filesIs the connection to company systems properly protected?
3. ApplicationAccounting software or an integration reads the general ledgerDoes the application have only the permissions it needs?
4. Data at restDocuments remain in shared folders and backupsWho can open the files, and how long should they remain?
5. PeopleA staff member emails a schedule to a client contactHas the recipient and attachment been checked?
6. AI model training and servicesA worker pastes a ledger into an AI toolWhat happens to the data after submission, including any training use?

The six points are a practical teaching map, not a formal classification prescribed by the PDPC or CSA. A single file may pass through more than one point during the same task.

1. Devices: the first local copy

Client files often arrive on a device before anyone thinks about storage policy. A laptop may hold downloaded attachments, desktop copies and synchronised folders.

A lost or compromised device can expose those copies. A sensible starting point is a device lock, software updates, appropriate malware protection and encryption where supported. Avoid transferring client records to an unapproved personal device simply to finish a task from home.

The CSA's data-breach advisory recommends measures including software updates, encryption of sensitive data and restricted access.

2. Networks: connecting to the records

A network carries information between devices and services. An accountant may work in the office, at a client's premises or remotely.

Do not assume that a connection is safe merely because an application opens. Use the firm's approved remote-access method, protect the account and avoid bypassing required security settings. The technical design belongs with the firm's IT provider or administrator, but every user should know which connection method is approved.

3. Applications: more permission than necessary

Applications are useful because they can import transactions, retrieve documents and exchange data. Some integrations continue to have access after their original task has ended.

Before approving a connection, ask what the application may read, change, download or share. An integration that only needs to read a report should not automatically receive permission to modify all records.

Access should be reviewed when staff leave, roles change or a service is no longer used. Restricting unnecessary access reduces the information available if an account or integration is compromised. This is consistent with the access-control principles described by ISO/IEC 27002:2022. The standard is guidance for information-security controls, not a law that automatically applies to all Singapore accounting firms.

4. Data at rest: stored files and backups

A completed assignment may leave copies in email, working folders, shared drives and backups. Keeping information for a legitimate purpose is different from allowing everyone continued access.

Review shared-link settings, folder memberships and retention requirements. Protect backups and check that important records can be recovered. The CSA's data resilience advisory recommends regular backups and consideration of more than one backup option for important information.

A retention decision should reflect applicable legal, professional and client requirements. Do not delete records solely because an engagement is finished.

5. People: a simple mistake can cross the boundary

Human handling includes sending documents, granting access, copying information and following requests. A wrong recipient can receive a genuine confidential file without an attacker breaking into any system.

Before sending a sensitive attachment, check the recipient, the document and the intended purpose. Where possible, use approved secure sharing with named recipients rather than public links. Make it easy for staff to report a mistaken send promptly so the firm can contain the problem.

The same rule applies to financial instructions. The Singapore Police Force's 2026 business email compromise advisory and the FBI's guidance both recommend verifying unusual changes through an independent channel. A genuine-looking email thread is not adequate evidence of a changed bank account.

6. AI services: what happens after submission?

An accountant may use an AI assistant to summarise a ledger or draft a client explanation. That can involve transferring confidential information to another service.

The central issue is what the provider is allowed and able to do with the submitted data. Training use is one possible consideration, but not the only one. Review the provider's applicable terms and settings, retention arrangements, access permissions, sharing or subprocessors, and the firm's client obligations. Do not assume that every AI product trains on submitted data, or that a setting available in one product applies to another.

Where the task does not require identifiable records, consider removing unnecessary identifiers before using the tool. Even then, the firm should check whether the remaining details could identify the client. The CSA Guidelines on Securing AI Systems support risk-based controls for AI use.

What should you do with this map?

Choose one real workflow and trace a document through all six points. Record where it arrives, which people and applications can reach it, where copies remain and when external services receive it.

Then ask three questions at each point: Who needs access? What could go wrong? Which control reduces the risk? Do not assume every problem needs a new product. Often the first improvement is a clearer permission, a recipient check or a documented approval.

Key takeaways

  • Exposure is possible through devices, networks, applications, storage, people and AI services.
  • A legitimate reason to process client data does not justify unrestricted access or disclosure.
  • Mapping one workflow provides a practical starting point for improving controls.

Related reading: Why Accounting Firms Are Targets explains why these records and access paths matter.

Last source check: 9 November 2026.

Next in the course: Recognising Phishing and Fake Regulator Messages

Sources

Frequently asked questions

This information has been prepared for general informational purposes only and is not intended to be relied upon as accounting, tax, or other professional advice.

Related articles