The Rules Are Two Documents, Not Twenty
Outbound calling compliance in Australia sounds like a thicket and is actually two instruments, each answering one question.
| Instrument | Question it answers |
|---|---|
| Do Not Call Register Act 2006 (Cth) | Who may you call? Which numbers can be listed, and the washing mechanism that keeps you clear |
| Telecommunications (Telemarketing and Research Calls) Industry Standard 2017 | When and how? Hours, the number you present, what you must say, and when you must stop |
The ACMA monitors compliance and the Register has been operating since May 2007. This is not emerging regulation. It is a settled framework that many businesses have never read, because outbound calling tends to begin as an informal activity rather than a project with a compliance step in it.
Every breach described below is a settings problem or a training problem. Neither requires a change of heart — which is why all four have permanent fixes.
The premise of this article
Breach One: The Dialler That Kept Going
The most common breach in Australia, and the least deliberate. A campaign is behind target, the team pushes on, and calls go out at 8:10pm on a Thursday.
The permitted hours under the Industry Standard:
| Day | Telemarketing | Research |
|---|---|---|
| Monday–Friday | 9:00am – 8:00pm | 9:00am – 8:30pm |
| Saturday | 9:00am – 5:00pm | 9:00am – 5:00pm |
| Sunday | No calls | 9:00am – 5:00pm |
| Public holidays | No calls | No calls |
But knowing the hours is not the hard part. Three things make this breach happen to people who know the table perfectly well.
The hours belong to the person you call
Not to your office. Dialling Sydney from Perth at 5:15pm AWST reaches someone at 7:15pm. Dialling Perth from Sydney at 9:15am AEST reaches someone at 6:15am, three hours before the window opens. Both directions breach, and daylight saving makes it worse for half the year.
Public holidays are not national
Australia has state and regional holidays. A dialler that only knows national ones will call Melbourne on the Friday before the Grand Final and Brisbane during the Ekka. The calendar has to be per state.
The last minute is the risky one
A dialler placing attempts until exactly 8:00pm will place some after it. There is no margin in running to the legal limit.
Calling it research does not make it research
What the call does determines which column applies. A survey that ends with an offer is telemarketing, and naming it research does not buy the Sunday window.
The fix that holds
Dial windows configured per state, derived from the destination number rather than your own location, with a per-state holiday calendar and an operational cutoff before the legal one. Stop at 7:45pm, not 8:00pm. Configured this way the system will not place the call, so the campaign being behind target cannot produce a breach — which is the point, because the pressure that causes this breach is real and will recur.
Breach Two: The List Somebody Bought
Second most common, and the one with the least awareness attached. A list arrives from a data vendor, described as clean, and calling begins.
The Register works by prohibition: calling a listed number is not permitted. So the first thing to know is which numbers can be listed.
| Registrable? | Type of number |
|---|---|
| Yes | Numbers used for private or domestic purposes — household lines and personal mobiles |
| Yes | Numbers used exclusively by government bodies, and emergency service numbers |
| Yes | Numbers used exclusively for faxes |
| No | General business telephone numbers used for ordinary operations |
This is where B2B callers relax, and where they should not. Yes, a company’s published main line cannot be on the Register. But an enormous number of Australian businesses run entirely on a personal mobile, and a number used for private or domestic purposes can be registered no matter what directory it appears in. A B2B list of small businesses is substantially a list of personal mobiles.
Washing is the mechanism that resolves this. You submit your list, it comes back marked, and you get a defined reliance period.
30 days
Reliance on a washed list, from when it was returned to you
30 days
Time marketers have to action a newly registered number
May 2007
When the Register started operating
4
Ways a wash still fails to protect you — below
Four ways a wash fails to protect you
You logged the wrong date. The 30 days run from when the washed list came back, not from when you sent it. Log the submission date and you overstate your window by however long the wash took.
The campaign outlived the wash. A six-week campaign on one wash is uncovered for the last fortnight. Set a standing monthly wash instead of washing per campaign.
You washed a different file. Lists get merged, enriched and appended. The file that matters is the one the dialler loads, not the version you washed three edits ago.
You trusted the vendor. “DNC cleaned” on an invoice is not your compliance. You need a wash you can date and produce, tied to the numbers you actually dialled.
And separately from the Register: maintain your own permanent suppression list. Someone who tells you not to call again must be suppressed by you, forever, regardless of what the Register says.
Breach Three: The Number That Cannot Be Called Back
Usually inherited rather than chosen. A trunk or dialler was configured years ago to withhold the outbound number, or the campaign presents a number that rings out.
The Standard is explicit: calling line identification must be enabled at the time you make or attempt to make a call. Withheld numbers on outbound campaigns are not permitted. And a second obligation rides with it: a return contact number must remain available for at least 30 days after the original call.
| Obligation | What goes wrong |
|---|---|
| CLI enabled on every call and attempt | A default setting nobody has reviewed since the system was installed |
| The presented number is genuinely reachable | An outbound-only number that rings out, or has no voicemail and no queue behind it |
| Reachable for at least 30 days after the call | Campaign numbers switched off the week the campaign finishes |
The fix, and why it is also better commercially
Present one permanent, answered business number on all outbound calling, with a queue or at least voicemail behind it, and lock it at the platform level so it cannot be withheld or overridden per campaign. That satisfies all three obligations at once. Rotating numbers per campaign creates a fresh 30-day commitment on each one, and rotating numbers to avoid recognition is precisely the pattern the ACMA's scam-disruption work targets. If you need callbacks routed to different teams, route on the number that was dialled instead of using disposable numbers — see 1300 and 1800 numbers.
Worth noting how completely compliance and effectiveness align here. Australians answer very few calls from unknown numbers in 2026. A withheld number is not just a breach, it is a campaign that does not connect. A permanent, recognisable, answered number is the only version of this that works at all.
Breach Four: The Rebuttal After the No
The one that is a training problem rather than a settings problem, and the hardest to fix for exactly that reason.
The Standard requires that a caller terminate the call immediately if the recipient asks, or indicates that they do not wish to continue.
Read the word indicates carefully, because it is doing a lot of work. The obligation is not confined to a person reciting a particular formula. If somebody signals they do not want to continue the call, the call ends.
This directly contradicts standard sales training
Objection handling teaches the opposite reflex: that the first no is not a real no, that a rebuttal should be attempted, that a supervisor might save the call. In an Australian outbound context, after a clear indication that someone wants to stop, all of those are non-compliant. No rebuttal script. No “before you go, can I just ask”. No warm transfer to someone more persuasive. This has to be trained explicitly and reinforced, because the instinct you are overriding is the one your salespeople were rewarded for developing.
The operational fix has two halves.
- Make suppression one click. A single action on the agent’s screen that writes the number to a permanent do-not-contact list, checked automatically before any future dial. If honouring an opt-out requires filling in a form later, it will not happen reliably.
- Check it on every call, not a sample. Traditional quality assurance reviews a tiny fraction of calls, so a rebuttal-after-no habit can run for months undetected. Automated call scoring can check every call for whether the agent stopped when asked — see AI call scoring and quality assurance.
What You Have to Say in the First Ten Seconds
Not a breach category of its own so much as an omission, and it is nearly universal.
| Call type | Required information |
|---|---|
| Telemarketing | The employer’s name, the purpose of the call, and who authorised the call |
| Research | The employer’s name and the purpose of the call |
The element almost always missing is the third one for telemarketing: who authorised the call. It matters most in the arrangement where it is most often skipped — an agency or contact centre dialling on behalf of a client brand. The person on the phone has to be able to say on whose behalf they are calling, not just who employs them.
Build all of it into the opening rather than making it contingent on the recipient asking. And once again the commercial and compliance interests coincide: the three seconds of ambiguity at the start of an unidentified call is exactly when Australians hang up.
The Exemption That Is Narrower Than You Think
Some callers and call types sit outside the Register’s telemarketing prohibition — broadly, categories treated as serving a public interest: government bodies, charities and charitable institutions, religious organisations, educational institutions contacting their own communities, and political parties, representatives and candidates.
An exemption from the Register is not an exemption from the Standard
Permitted hours, calling line identification, the required disclosures and the obligation to stop when asked all continue to apply. A charity calling donors at 9pm on a Sunday from a withheld number is not compliant merely because the Register does not prohibit it from calling. If you work in one of these sectors this is the paragraph to take away, because the exemption is routinely understood as much broader than it is.
Two more boundaries that come up constantly in ordinary commercial life:
An existing customer is not a blanket exemption. Whether a call is a telemarketing call depends on what the call is doing. Confirming an appointment, advising a delivery or following up a fault is a service call. Ringing to sell an additional product is telemarketing, however many years someone has been a customer.
Consent does not override the Standard. Consent addresses whether you may contact someone. It does not extend your calling hours, permit a withheld number, or excuse you from identifying yourself. The Standard governs how the call is conducted regardless of how you obtained the number.
When the Caller Is an AI
The live question of 2026, with a straightforward answer: the rules attach to the call, not to whether a human is speaking.
An AI voice agent placing an outbound call is making a call. The Register governs who it may call. The Standard governs the hours, the CLI, the disclosures and the obligation to stop. Neither instrument turns on what is producing the voice.
What genuinely changes is the risk profile, in three ways.
Scale multiplies errors
A human making a configuration-driven error makes it a few dozen times and probably notices it is dark outside. An automated system executes a misconfigured schedule perfectly, thousands of times, without hesitation. Test the configuration against the rules before volume, not after the complaints.
Stopping is the hard requirement
The obligation triggers when someone indicates they do not wish to continue, not on an exact phrase. An agent has to recognise a wide range of refusals — including sighs, interruptions and "look, I'm not interested" — and end the call rather than attempt to re-engage. This is the capability to test hardest before going live.
Evidence gets easier
The upside. Automated calling produces complete logs by default: every attempt timestamped, every opt-out recorded, every disclosure delivered identically. Used properly that turns a complaint into a two-minute answer instead of a reconstruction.
Two further points worth stating plainly. Say that it is automated. No rule on this page compels that disclosure, but people react badly to discovering mid-call that they have been talking to a system, and the transparency direction of Australian AI policy runs one way. And more fundamentally: outbound is the worst place to start with AI calling. The goodwill is thinnest and the reputational downside largest. Answering calls people chose to make is a far safer first project — which calls to automate and AI voice agent cost and ROI both make that case.
Make the Platform Do It
The thread running through all four breaches: each one can be prevented by careful people remembering things, and that approach fails eventually. People work late, campaigns get handed over, and the person who knew the rules leaves.
Almost all of it can be enforced by configuration instead, where it holds without anyone thinking about it.
| Rule | Configuration that enforces it |
|---|---|
| Permitted hours | Dial windows per state, derived from the destination number, with an operational cutoff earlier than the legal one |
| Public holidays | Per-state holiday calendar maintained annually, blocking outbound campaigns automatically |
| CLI enabled | Presented number locked at platform level, not overridable per campaign or user |
| Reachable return number for 30 days | One permanent business number with a queue behind it. No disposable campaign numbers |
| Stop on request | One-click suppression writing to a permanent list, checked before every dial |
| Washed within 30 days | Standing monthly wash of the file the dialler loads, return date recorded, washed file retained |
| Required disclosures | Scripted into the first ten seconds and checked on every call by automated scoring, not a 2% sample |
| Evidence if challenged | Timestamped call logs, opt-out records and washed-list history retained together |
If you are choosing a platform to run outbound on, those eight rows are the specification. Contact centre software in Australia covers what else to look for.
Legal Is Not the Same as Welcome
One honest closing point, because an article about compliance can leave the wrong impression.
Everything above describes the floor. It is entirely possible to run a campaign that satisfies the Act and the Standard in every respect and is still an unwelcome interruption that damages your brand. The rules do not require anyone to be pleased that you called.
The convenient part
Notice that every single fix in this article also improves results. A permanent, answered number gets called back. A sensible hour reaches someone able to talk. Identifying yourself in the first five seconds beats the ambiguity that makes people hang up. Honouring opt-outs stops you burning list you paid for and generating complaints that cost more than the deal. There is no trade-off here between compliant and effective — which is rare enough in regulation that it is worth taking full advantage of.
The businesses doing well on the phone in Australia right now are mostly not the ones with the best-tuned diallers. They are the ones getting called back, because the number was real, someone answered it properly, and the reason for the original call made sense to the person receiving it. Configure the compliance so it stops being a thing you think about, then spend the attention on having a reason to call.
Frequently Asked Questions
What hours can we make sales calls in Australia?
Telemarketing calls are permitted Monday to Friday from 9:00am to 8:00pm and Saturday from 9:00am to 5:00pm, with no calls on Sundays or public holidays. Research calls have different hours: weekdays 9:00am to 8:30pm, Saturday 9:00am to 5:00pm, and Sunday 9:00am to 5:00pm, with no calls on public holidays. Knowing the table is the easy part. Three things cause breaches among people who know it perfectly well. The hours belong to the person being called rather than to your office, so dialling Sydney from Perth at 5:15pm local time reaches someone at 7:15pm, and dialling Perth from Sydney at 9:15am reaches someone at 6:15am, three hours before the window opens. Public holidays are state and regional rather than national, so a dialler that only knows national holidays will call Melbourne and Brisbane on days it should not. And a dialler running to exactly 8:00pm will place some attempts after it, so set an operational cutoff around 7:45pm. Finally, whether a call is telemarketing or research depends on what the call does, so a survey that ends with an offer is telemarketing and does not get the Sunday window.
We only call businesses. Does the Do Not Call Register apply to us?
Partly, and the gap is where B2B callers get caught. General business telephone numbers used for ordinary operations cannot be registered, so calling a company's published main line is not prohibited by the Register. But an enormous number of Australian businesses run entirely on a personal mobile, and a number used for private or domestic purposes can be registered regardless of which directory it appears in, which means a B2B list of small businesses is substantially a list of personal mobiles. More importantly, the Telecommunications (Telemarketing and Research Calls) Industry Standard 2017 applies to your calls whether or not the Register does. Permitted hours, calling line identification, the required disclosures and the obligation to stop when asked are all independent of the Register. The mistake that causes real trouble is establishing that B2B calling is permitted and concluding that nothing else applies. The practical advice is to wash your list anyway, because it is inexpensive and it removes precisely the sole-trader mobiles that carry the actual risk.
How does list washing work and how long is a wash good for?
You submit your calling list to the Register operator and it comes back marked with which numbers are listed. If you washed within the 30 days before making a call and the Register did not indicate that number was listed, you are not in breach for calling it. Four things stop a wash protecting you. First, the 30 days run from when the washed list was returned to you, not from when you submitted it, so logging the submission date overstates your window by however long the wash took and leaves the tail of a campaign uncovered. Second, a campaign can outlive its wash, so a six-week campaign on a single wash is unprotected for the last fortnight; set a standing monthly wash instead. Third, lists get merged, enriched and appended, and the only file that matters is the one the dialler actually loads, not the version you washed three edits ago. Fourth, a vendor's assurance that a list is DNC cleaned is not your compliance; you need a wash you can date and produce, tied to the numbers you dialled. Separately, keep your own permanent suppression list, because anyone who asks not to be called again must be suppressed by you forever regardless of the Register.
Do we have to show our phone number on outbound calls?
Yes. Calling line identification must be enabled at the time you make or attempt to make a call, so withheld or blocked numbers on outbound campaigns are not permitted. A second obligation travels with it: a return contact number must remain available for at least 30 days from the original call, which means the number you present has to be genuinely reachable and must stay live and answered for a month afterwards. The three usual failures are a default setting nobody has reviewed since installation, presenting an outbound-only number that rings out with no voicemail or queue behind it, and switching off campaign numbers the week the campaign ends. The fix that satisfies all of it is to present one permanent, answered business number on all outbound calling with a queue behind it, locked at the platform level so it cannot be withheld or overridden per campaign. Rotating numbers creates a fresh 30-day commitment on each one, and rotating them to avoid recognition is exactly the pattern the ACMA's scam-disruption work targets. Route callbacks on the number that was dialled rather than using disposable numbers.
If someone says they are not interested, can our salesperson try one more time?
No, and this is the requirement that conflicts most directly with ordinary sales training. The Standard requires a caller to terminate the call immediately if the recipient asks, or indicates that they do not wish to continue. The word indicates matters: the obligation is not confined to someone reciting a particular phrase, so if a person signals they do not want to continue, the call ends. That rules out the rebuttal script, the one-more-question gambit, and the warm transfer to a more persuasive colleague. Objection handling teaches the opposite reflex, so this has to be trained explicitly and reinforced, because you are overriding the instinct your salespeople have been rewarded for developing. Two operational supports make it stick. Make suppression a single click on the agent's screen that writes to a permanent do-not-contact list checked before any future dial, because an opt-out requiring a form later will not happen reliably. And check compliance on every call rather than the tiny sample traditional quality assurance reviews, since a rebuttal-after-no habit can otherwise run undetected for months.
What must we say at the start of a telemarketing call?
For a telemarketing call you must provide the name of your employer, the purpose of the call, and who authorised the call. For a research call you must provide the employer's name and the purpose. The element almost always missing is the third one for telemarketing, and it matters most in exactly the arrangement where it is most often skipped: an agency or contact centre dialling on behalf of a client brand. The person on the phone needs to be able to say on whose behalf they are calling, not merely who employs them. Build all of it into the opening rather than making it conditional on the recipient asking. This is another case where compliance and effectiveness point the same way, because the few seconds of ambiguity at the start of an unidentified call is precisely when Australians hang up.
Do these rules apply to AI voice agents making outbound calls?
Yes. The rules attach to the call rather than to whether a human is speaking, so an AI agent placing an outbound call is subject to the Register on who it may call and to the Industry Standard on hours, calling line identification, disclosures and stopping when asked. Neither instrument turns on what produces the voice. What changes is the risk profile. Scale multiplies configuration errors, because a human would notice it was dark outside while an automated system executes a misconfigured schedule perfectly thousands of times, so test configuration against the rules before volume rather than after complaints. Stopping on request is the hardest capability to implement, since the obligation triggers when someone indicates they do not wish to continue rather than on an exact phrase, so the agent must recognise a wide range of refusals and end the call rather than re-engage. The upside is evidence, since automated calling produces complete timestamped logs and opt-out records by default. It is also worth disclosing that the caller is automated, not because a rule compels it but because people react badly to discovering it mid-call. More fundamentally, outbound is the worst place to start with AI calling, because the goodwill is thinnest and the reputational downside largest; answering calls people chose to make is a far safer first project.
