AI Tax Research Software and Client Information: Do You Really Know What’s Happening to Your Data?

“Before you place client information into any LLM, tax research software, or agentic AI process that harnesses LLM technology, you must ask yourself: ‘Do I actually know what is happening to the data?’” says McPhedran.

“This is a direct requirement of the Tax Practitioners Board, set out in TPB(I) D62/2026, released on 23 March 2026. If you do not know exactly what happens to the data, you cannot possibly be in a position to protect it or obtain the correct client permissions. We’ve noticed the data security angle becoming a major focus since this message has spread, quite rapidly, through the profession.”

For a time, the prevailing view among practitioners was that an enterprise-grade LLM subscription provided sufficient protection. Many accounting firms are now realising that is no longer the case. The United States CLOUD Act permits US law enforcement to compel data from any US-controlled provider — regardless of where that data is physically stored. This means client data processed through platforms like Microsoft Copilot, ChatGPT, or Claude may be accessible to US authorities at any time, without your knowledge or your client’s. You cannot input sensitive client information into an overseas AI platform without specific, explicit client consent — consent that, in most standard engagement letters, simply does not exist. For firms without that consent in place, the exposure under TPB(I) D62/2026 and the Privacy Act 1988 is real and growing.

A Local Answer to a Global Problem

Elfworks, the Brisbane-based, UniQuest Extension Fund-backed AI accounting platform now has over 500 firms on the platform, including 15 of Australia’s Top 50 accounting firms as paying customers. with strong growth continuing through word-of-mouth alone. While productivity and accuracy were the initial drawcard, the key point of difference for Elfworks is rapidly becoming the security of client information, underpinned by the platform’s three-tiered approach to data protection.

A Three-Tiered Defence

Elfworks Managing Director Ian Youngman, whose background spans digital transformation and data security, is adamant that robust data security is the absolute starting point for any serious AI provider in this space.

“Data security was our very first concern when we designed Elfworks,” Youngman explains. “We developed a three-tiered approach:”

1. Client Data Anonymisation. Sensitive client data is systematically redacted prior to LLM processing and securely re-populated upon completion — meaning even if data were compelled under foreign law, there is nothing identifiable to hand over.
2. Sovereign Language Models. We utilise Australian-hosted AI models managed by Australian companies. They do not store your data, do not train on your data, and are not obligated to provide data to any foreign government or agency.
3. Enterprise-Grade Frontier Models. For tasks requiring the deepest knowledge base, we use enterprise-grade subscriptions to leading global LLMs — but only after client data has been anonymised.

“This is the core of the Elfworks approach — AI models working together, selecting the right model for the right level of risk, with uncompromising data security as the foundation.”

This security-first architecture is paying off. “We are noticing a distinct shift where data security has become the number-one priority for partners at Australia’s leading firms. That is why we continue to build layers of protection around client data,” says Youngman.

Putting Capital Where It Counts

Backed by a successful capital raise earlier this year, Elfworks has been able to channel significant resources directly into its security technology stack.

“Our marketing spend to date has been less than budgeted, due to incredibly strong word-of-mouth growth,” Youngman says. “This has freed up funds to invest heavily in our data security infrastructure. Our anonymisation technology and sovereign model architecture has allowed us to safely release our first wave of Elfworks agents.”

These specialised agents — including the Client Structure Builder, Trust Resolution Drafter, and Year-End Tax Planner — perform complex, linked tasks using sensitive client information that remains safely within Australian borders.

“Because we use advanced anonymisation and sovereign, Australian-hosted models, client data does not leave our jurisdiction. Accountants can use these tools without needing client permission for data to be stored overseas or to be subject to foreign laws that could compel disclosure.”

The Bottom Line for Australian Practitioners

As the TPB transitions from consultation to enforcement of its new AI guidelines, the stakes for Australian accounting firms have never been higher.

“It’s only a matter of time before the Tax Practitioners Board starts to actively hold accountants to the standards in TPB(I) D62/2026,” warns McPhedran.

“Ultimately, it all starts with one simple question: Do you really know what happens to client information when you click submit?”

See Elfworks in action. Visit Elfworks.ai for a free trial.