Finsure Declined 13 Broker Applicants in Sentinel’s First Six Weeks. ASIC Told Bank Boards It Will Review Lender Oversight of Brokers From Q3

On Thursday 22 October, Australian Broker reported that Finsure had turned away 13 prospective brokers in the first six weeks that its in-house artificial intelligence risk system, Sentinel, had been running. According to that report, Sentinel identified association risks in every one of those 13 cases — risks the aggregator said a standard, surface-level check may not have picked up.

Three weeks before that, on 30 September, ASIC Commissioner Kate O’Rourke signed a supervisory letter to bank boards and executives setting out the regulator’s 2026–27 banking priorities. One section is headed Lending practices and broker oversight, with an expected commencement of Q3 2026–27. The letter says the work “is likely to focus on the impact of the short-term variable remuneration changes for proprietary lenders” and on “the use of referrers by lenders, and lender oversight of brokers”. It adds that the review “follows ASIC’s work with the Australian Prudential Regulation Authority (APRA) and AUSTRAC in relation to reported instances of mortgage loan fraud”.

Nobody has connected those two events, and this article does not. But put side by side, they describe a shift worth planning around. For most of the decade since the Royal Commission, the broker channel has been checked after the event: file reviews, audit samples, complaints data, best-interests assessments on recommendations already made. Both of these developments point somewhere else — to the gate. To the moment you apply, get accredited, or get re-accredited.

What an aggregator reported screening for

According to Australian Broker‘s report, Sentinel combines multiple data points to map connections that may point to past misconduct. The checks are described as covering an applicant’s track record, identity and conflicts of interest, plus the people and entities the applicant is linked to — including known problematic entities, undisclosed associates, and referral partners such as accountants. Assessments are said to follow set procedures, with the aim of catching problems at onboarding rather than later.

That last detail is the one brokers should sit with. A referring accountant is not a credit representative, is not on anyone’s panel, and does not appear on a loan file as a party to the transaction. But a referral relationship is a relationship, and relationships are exactly what a mapping system is built to find.

Bednar also described the system as “precision intelligence that augments better decision-making for our people, from the outset”, and said screening out bad actors protects legitimate brokers, who could otherwise be drawn into investigations through indirect associations. The report notes he said Sentinel is not designed to flag every broker with a remote link to a questionable party. That qualifier matters, and it is worth repeating before anyone starts auditing their contact list in a panic.

None of this appeared from nowhere. In February, The Adviser reported that Finsure had lifted its entry requirements: new members needed at least two years’ experience in loan assessment or in direct customer conversations assessing credit requirements, with applicants lacking that experience required to complete the aggregator’s Academy program. Sentinel was described at the time as an AI-driven monitoring tool being rolled out to flag potentially improper activity in the network. The October report is the first public account of what it has actually produced at the application stage.

Why a decline is not a finding

The same discipline applies to the wider backdrop. In August, the ABC reported that AUSTRAC’s Operation Claw had identified suspected mortgage fraud worth hundreds of millions of dollars across 10 major banks, concentrated in Sydney properties, involving allegedly inflated incomes, misrepresented employment and fabricated or unverifiable business activity. AUSTRAC said it had made referrals to appropriate authorities, and also said the project did not find evidence of widespread money laundering. Australian Broker‘s October report puts those referrals in the hundreds across brokers, lawyers, accountants and high-risk companies.

A referral to an authority is the start of a process, not the end of one. It is not a charge, and it is certainly not a finding of guilt. The reason it still matters to a broker who has done nothing wrong is mechanical rather than moral: once relationship data exists at that scale, screening on it becomes cheap, and the cost of being adjacent to it shifts onto the person who has to explain the adjacency.

The second gate: ASIC’s letter to bank boards

It is worth being precise about what ASIC’s September letter does and does not say, because the trade commentary around it has run ahead of the document.

What the letter says is quoted above: a review of lender conduct, expected to start in Q3 2026–27, likely to focus on proprietary-lender remuneration changes, the use of referrers by lenders, and lender oversight of brokers. It is addressed to bank boards and executives. The obligations it concerns are the lenders’.

What the letter does not say is anything about what lenders should require of individual brokers, or about relationship mapping, screening tools, or accreditation criteria. The Adviser‘s coverage on 30 September suggested the review could examine matters such as accreditation and re-accreditation, application quality, exception reporting and fraud indicators — but that is the outlet’s reading of the scope, not ASIC’s list, and the letter itself does not break the heading down that way.

Why this lands at onboarding, not in a BID file review

Brokers are used to thinking about scrutiny through the best interests duty, and ASIC’s guidance in Regulatory Guide 273 is the reference point most licensees build their processes around. But it is worth separating the two questions, because conflating them leads brokers to look for this risk in the wrong place.

The best interests duty is assessed against the credit assistance provided to a particular consumer: what was recommended, what was considered, what was recorded, whether the outcome served that client. It is a file-level question, answered from the file.

Association risk is not a file-level question. It asks who a licensee or a lender is prepared to contract with and authorise in the first place. That is why it surfaces at the gate rather than in an audit sample, and why a spotless record of best-interests documentation does not answer it. A broker can have five years of impeccable file notes and still face a question about a former business partner or a recurring referral source — because the question is not about the files at all.

This article is general information rather than advice on your own obligations. What any of this requires of you specifically depends on your licensing arrangement, and that is a conversation for your licensee, aggregator compliance team or your own compliance adviser — not for an article.

What association risk looks like in an ordinary book

The unhelpful version of this topic imagines a broker with something to hide. The useful version imagines a broker with nothing to hide and nothing written down. In a normal, honest book, the connections a mapping system would surface are almost entirely mundane:

• A referring accountant who also refers to a broker you have never met and know nothing about.
• A former co-director of a trading entity you wound up years ago, which still sits against your name on a company search.
• A buyer’s agent who appears across a cluster of your settled files in one corridor, for the ordinary reason that they work that corridor.
• A mentoring relationship, a shared office, a joint marketing arrangement, a white-label referral that ran for one quarter in 2023.
• A referral partner you quietly stopped dealing with because something felt wrong — and never recorded that you stopped, or why.

That last one is the expensive one. The difference between an association you can explain and one you cannot is almost always whether you documented it while it was current. A note written at the time reads as judgement. The same explanation reconstructed two years later, after someone else has raised it, reads as a defence.

There is a second reason to care, and it has nothing to do with screening. AUSTRAC’s chief executive, Brendan Thomas, told the ABC in August that “the most effective way to stop mortgage fraud is before a loan is approved”. Whatever a broker thinks about where oversight should sit, that sentence is the logic of every gate being built right now — and the broker channel’s own interest in clean referral pathways is identical to the regulator’s.

What to review this week

None of this needs a project. It needs about ninety minutes and a willingness to write things down before anyone asks for them.

1. Build a one-page referral register. Every source that has sent you business in the past 24 months: name, business, type of arrangement, whether anything is paid in either direction, and roughly how many files came from them. Not a CRM report — a page you could hand to someone.
2. Search yourself the way a screening process would. Your own name, your trading names, your current and former entities, and any directorships still recorded against you. Find the stale ones now, while discovering them is free.
3. Write the exit notes you never wrote. For every referral relationship you have ended, record what it was, when it ended, and the reason, in a sentence. Date it today and note that it is a retrospective record — a dated reconstruction is honest; an undated one is not.
4. Separate the referral from the money in your records. Where a referral arrangement involves payment, make sure the arrangement, the disclosure and the payment are documented as three things, not inferred from a bank statement.
5. Decide your answer to the awkward question in advance. If an aggregator or a lender asked you to explain a connection, which one would it be? You almost certainly know. Draft the two-sentence answer now, while it is a thought exercise rather than a response.
6. If you are planning to move, sequence it deliberately. Do steps one to five before you lodge an application, not after a request for information arrives mid-process.

A scenario: the broker planning a move

Consider a sole trader writing a steady book, planning to change aggregators early next year. Nothing in her history is remarkable. She has one wound-up company from a short-lived asset finance venture, a long-standing referral relationship with an accountant who sends her four or five files a year, and a buyer’s agent she stopped working with in 2024 after a valuation dispute she was not comfortable with.

Under a file-based model, none of that ever comes up: her files are clean, her notes are good, and the questions she gets asked are about volumes and lender mix. Under a gate-based model, two of the three are visible before she writes a single file on the new panel, and the third — the buyer’s agent she walked away from — is the one she can least easily evidence, because her reason for ending it exists only in her memory.

Her position is not weak. It is undocumented, which is a different problem with a cheaper fix. A dated file note written this week, saying what happened and why she ended the relationship, converts a vulnerability into the single strongest thing she could have: contemporaneous evidence that she exercised judgement without being asked to.

What to watch next

• Whether other aggregators say anything. Finsure is, so far, the only one to have publicly described screening of this kind and reported an outcome from it. Whether that stays unusual or becomes standard is the signal to watch over the next two quarters.
• What ASIC’s Q3 review actually asks banks. The letter names referrer use and lender oversight of brokers. How those translate into questions put to lenders — and from lenders into requests put to accredited brokers — is where the practical effect on your week will come from.
• Information sharing between institutions. Finsure said it intends to upgrade Sentinel’s detection capabilities and integrate it with industry intelligence frameworks, and that it is taking part in industry discussions on information sharing. A screening tool inside one aggregator is one thing; shared intelligence across several is a materially different proposition for brokers, and worth following closely.