AI and Professional Obligations
How private AI relates to the duties that bind an accounting firm: confidentiality, anti-money-laundering requirements, independence, data protection and retention. Technology can support these duties. The firm and its accountants remain responsible for meeting them.
1Confidentiality without a general privilege
Secure Accounting runs on premises for confidential client work. By default, authorised files are processed under the firm's control rather than being sent to a public or third-party AI model.
The accountant's legal position differs from the lawyer's.
In Malaysia and Singapore, accountants do not have a general accountant-client privilege equivalent to legal professional privilege. The confidentiality duty in the MIA By-Laws and the international ethics code is binding, but it does not defeat a lawful demand for information.
Control over where client information is processed is therefore an important part of confidentiality. It does not replace the firm's legal and professional controls.
Many firms already store files in Microsoft 365, Google Workspace or another cloud service. Cloud storage is not the same as sending those files to an external AI model for processing. Secure Accounting keeps its default AI processing on premises. If the firm uses optional Frontier Mode, names and identifiers are replaced through reversible pseudonymisation, the accountant approves what is sent, and original identifiers are restored inside the firm.
2The anti-money-laundering regime
Anti-money-laundering requirements place particular limits on how sensitive information may be handled.
Under Malaysia's AMLA, accountants are reporting institutions when they carry out specified activities, such as managing client money or forming companies. Audit or tax compliance alone does not create that status. Where the regime applies, the firm must conduct customer due diligence and report suspicious transactions. Tipping-off restrictions limit disclosure about a report or contemplated report.
KYC material and information connected with a suspicion require strict handling. They should be excluded from any general AI workflow unless the firm has confirmed a lawful, controlled purpose and the applicable safeguards. The exclusion should be enforced through the firm's information classification, access rules and approved data sources, not left to staff memory.
KYC material may sit beside routine client work. The firm should design its workflow so that a classification mistake does not automatically expose restricted information to AI processing.
3Independence and the audit file
For audit-registered firms, AI touches independence and evidence directly. This is developed fully in C9 and only summarised here.
A self-review threat may arise when the same tool helps prepare a figure and is then relied upon for assurance over it. The audit file must remain inspectable. Any AI contribution to a conclusion should be traceable to its source and the accountant's review.
Both are governance decisions taken at engagement level, not settings on a machine.
4Data protection
Data-protection requirements in Malaysia and Singapore apply alongside professional duties.
Malaysia's Personal Data Protection Act, as amended in 2024, now places direct, penalty-backed obligations on data processors, requires a Data Protection Officer, imposes mandatory breach notification, and applies a risk-based cross-border-transfer regime. Singapore's PDPA carries an equivalent DPO requirement, breach-notification duty and transfer-limitation obligation.
The firm should limit AI processing to authorised engagement purposes and approved information. Access should be granted only to staff who need it. Sensitive personal data, trust-account records and other restricted categories require controls that reflect the applicable law and the firm's policy.
Keeping confidential AI processing on premises reduces the need to send client information to an external model. It does not remove every transfer question, particularly where the firm's source systems, backups or optional services involve other jurisdictions. The firm must assess its actual data flows.
5Retention and erasure, which pull in opposite directions
Retention and erasure requirements may point in different directions.
Audit working papers and tax records carry statutory retention periods. Data-subject rights under both PDPAs carry erasure obligations. These pull in opposite directions, and a firm needs to be able to say which prevails for a given item and why.
An AI system may create derived data such as indexed text, embeddings, cached context and provenance records. The firm's retention design should account for each of these, together with backups. Disposal should remove the relevant copies when erasure is lawful and required.
Technical support for erasure does not decide the legal retention question. It allows the firm to carry out a documented decision.
6Technology does not carry the duty
Secure Accounting supports the firm's controls. It does not take over the firm's obligations.
Requirements to appoint a Data Protection Officer, notify qualifying breaches, conduct customer due diligence, report suspicions, respect data-subject rights and exercise professional judgement continue to apply.
Secure Accounting is decision-support, not a decision-maker. AI can be wrong. Accountants must verify its work and remain responsible for every figure, filing and professional conclusion.
Compliance belongs to the firm. Technology can make controls visible and repeatable, but the firm must still conduct due diligence, maintain access rights, involve its Data Protection Officer and review relevant records.
Private AI is most useful when it operates within that governance, not in place of it.
Notes & further reading
- Confidentiality and the absence of privilege: the MIA By-Laws (On Professional Ethics, Conduct and Practice) adopt the international ethics code's confidentiality duty, which is professional and contractual rather than a legal privilege capable of resisting compulsion. Confirm with qualified counsel.
- Anti-money-laundering: Malaysia's AMLA 2001, covering accountants as reporting institutions when they carry out gazetted activities (not audit or tax-compliance work alone), customer due diligence, suspicious-transaction reporting and the tipping-off offence. See MIA, Obligations by Reporting Institutions Under AMLA.
- Data protection: Malaysia PDPA (Amendment) 2024 (processor obligations, mandatory DPO, breach notification, risk-based cross-border transfers) and the Cross-Border Personal Data Transfer Guidelines; Singapore PDPA equivalents.
- Retention: statutory retention periods for audit working papers and tax records under the applicable standards and the Income Tax Act; sequence disposal decisions against both retention and erasure duties.
- Independence and audit evidence are treated in full in C9.
This paper is a perspective for firm leadership and is not legal, tax or accounting advice. Professional and regulatory obligations differ by jurisdiction and continue to develop; firms should confirm specifics with qualified advisers before relying on them.