On 2 January 2020, Veronica Nelson died alone in a cell at the Dame Phyllis Frost Centre, Victoria's main women's prison, days after being refused bail and remanded for shoplifting-related charges. She asked prison staff for help through the night. The coroner who examined her death found, in January 2023, that it was preventable, and that the offences she was held for were exactly the kind of low-level, non-violent charges the bail system is not supposed to imprison anyone for before trial. Four years on, in a growth-corridor suburb an hour's drive away, a different household wakes to an empty driveway — one of the roughly one in five vehicles now stolen in Victoria using a relayed key-fob signal, often by offenders police describe as children. Both scenes are real, and both are Victoria's bail laws at work, in a state that has rewritten those laws twice in three years trying to answer both at once.
What the government did
The 2023 reform. Coroner Simon McGregor's findings into Veronica Nelson's death, handed down 30 January 2023 after a lengthy inquest, concluded it was preventable and that Victoria's bail system had "a discriminatory impact on Aboriginal people," producing grossly disproportionate rates of remand — worst of all for Aboriginal women. The Andrews government accepted the finding's thrust. The Bail Amendment Bill 2023 — nicknamed "Poccum's Law" after Ms Nelson's family name — was introduced 15 August 2023 and commenced 25 March 2024, by which time Jacinta Allan had succeeded Daniel Andrews as Premier. It repealed two bail offences that had criminalised breaching a bail condition or reoffending while on bail — the offences that had trapped Ms Nelson — required decision-makers to weigh an accused person's Aboriginality and its history of trauma and dispossession, and made it harder to remand people, including children, charged only with summary offences.
The age of responsibility backflip. A parallel 2023 commitment followed a similar arc. In April 2023 the government promised to raise the minimum age of criminal responsibility from 10 to 12 by the end of 2024, and to 14 by 2027. The Youth Justice Act 2024, passed in August 2024, lifted it to 12 from 2025 and wrote in a rebuttable presumption that children under 14 lack criminal capacity — but Premier Allan abandoned the promised further move to 14 in August 2024, citing community concern about crime. The government's own Commission for Children and Young People called the decision a step backwards, given the evidence the government had itself commissioned pointed the other way.
The reversal
Amid rising pressure over aggravated burglaries and car theft (below), the government introduced the Bail Amendment (Tough Bail) Bill on 18 March 2025 and passed it within days. It did more than adjust the 2023 settings — it inverted their philosophy: making community safety of overarching importance in every bail decision, removing the principle that remand should be a last resort for a child, creating new bail offences, and applying a strict test to serious offences including aggravated burglary, home invasion, carjacking and armed robbery. Electronic monitoring for young offenders expanded, and Australia's first statewide machete ban followed under the Control of Weapons Amendment (Machete Ban) Act 2025, passed May 2025 and effective 1 September 2025. In August 2025 a further Bail Amendment Bill extended the strict "high harm, high probability of reoffending" test to more repeat offenders. Then, in the first days of December 2025, Parliament passed the government's "Adult Time for Violent Crime" laws — modelled on Queensland's approach — allowing children aged 14 and over to face adult sentencing, up to life imprisonment, for offences such as aggravated home invasion and carjacking. Attorney-General Sonya Kilkenny told Parliament the laws were, in part, "incompatible with the human rights set out" in Victoria's own Charter of Human Rights and Responsibilities — a rare instance of a government legislating something it has formally acknowledged breaches its own human rights law.
The strongest case
A small number of children are doing outsized harm, and communities are entitled to safety
The government's defence has a real evidentiary base, not just a political one. Victoria Police intelligence and Crime Statistics Agency data both show a relatively small cohort of repeat young offenders drives a disproportionate share of the most frightening crimes: child and youth offenders accounted for close to half of aggravated burglary offenders in the year to March 2024, a year in which recorded aggravated burglaries rose 18.4% to 6,031 offences, while motor vehicle theft climbed to its highest level since 2002 — partly driven by keyless relay theft, with police estimating around one in five stolen vehicles taken that way. These are not victimless offences: aggravated burglary means an occupied home entered by force, sometimes at night, sometimes involving a weapon, and the trauma to households is real regardless of the offender's age. A justice system's first duty is to the people who did nothing wrong. Where a small number of children are before the courts again and again for high-harm offending, a bail system that keeps returning them to the community with no capacity to manage that risk is not protecting anyone, including, in the long run, the children themselves. Electronic monitoring, tighter tests for the worst repeat offences and faster consequences for weapon use are a defensible, evidence-informed response to a genuine shift in the offending pattern, not simply a reaction to headlines.
The evidence
Victoria's own data tells a more complicated story than either "crime is out of control" or "the reforms caused a crime wave" allows. The Crime Statistics Agency recorded 7,644 offenders aged 10 to 17 proceeded against by police in 2024-25 — a youth offender rate of 1,145 per 100,000 people in that age group, and, notably, a fall of roughly 6% on the year before (approx.). What has changed is the composition of that offending: a shrinking pool of distinct youth offenders is responsible for a rising share of the most serious offence types, consistent with the government's "small cohort of repeat offenders" framing — and with a reoffending pattern the state's own Sentencing Advisory Council has documented for a decade. Its 2016 study of children sentenced in the Children's Court found the highest reoffending rate of any age group was among 10-to-14-year-olds, at 81% over an 11-year tracking period, against 45% for the sentenced population as a whole — evidence the youngest offenders are also the least deterred by the court process itself.
On the remand side, the Council's separate study found two-thirds of children remanded in custody in 2017-18 did not go on to receive a custodial sentence — meaning most children held before trial spent time in detention for an outcome that, once their case was actually heard, did not warrant it. That is the exact pattern the Nelson finding was concerned with, and the pattern the 2025 reversal risks repeating at scale. The Productivity Commission's Report on Government Services shows Victoria's Aboriginal youth detention rate, at 8.1 per 10,000 in 2024-25, running well below the national rate of 25.7 per 10,000 — genuine evidence the 2023 reforms and a decade of the Aboriginal Justice Agreement had worked, delivering a 52.3% reduction in Aboriginal youth incarceration since 2019 against a 30%-by-2031 target. But the state's own Aboriginal Affairs reporting shows that progress reversing: the average nightly number of Aboriginal children in detention rose from 7.9 in the June quarter of 2024 to 12.9 in the June quarter of 2025, and 72.2% of those children were unsentenced in that most recent quarter — held on remand, not serving a sentence.
The critique
Judged against its own evidence base, Victoria's bail policy has whiplashed faster than the institutions built to inform it could keep pace with. The Bail Amendment Act 2023 was the direct legislative response to a coronial finding — arguably the most authoritative evidence a state government can act on, produced by an independent judicial officer after a multi-year inquest. It commenced in March 2024. Less than twelve months later, before the Sentencing Advisory Council or the Crime Statistics Agency had published a full year of data under the new settings, the government legislated in the opposite direction, explicitly removing the "remand as a last resort" principle the 2023 reform had been built around. A coroner's finding that survives less than a year before being substantially reversed suggests the reversal owed less to a considered evaluation of the 2023 reform than to the political weather — months of front-page coverage of carjackings and an opposition running hard on crime.
The fairness case for the 2025 laws is real, but it does not require abandoning everything the Nelson inquest found. High-harm repeat offending by a small cohort of children is a genuine problem the 2023 reforms, aimed mainly at low-level and summary offences, were never designed to solve — the government could have tightened bail for that cohort without reviving the opposite of the "last resort" principle for every child, and without an "adult time" sentencing model it has itself told Parliament breaches the state's own Charter of Human Rights. That deserves weight on its own terms: when an Attorney-General states on the record that a bill is "incompatible" with the human rights framework Parliament itself enacted, and the bill passes anyway, the ordinary safeguards against rushed law-making have not functioned as intended, whatever the underlying crime problem's seriousness.
A coroner's finding took two years and a woman's death to produce. Its central reform lasted less than one before the government moved to undo it.
The evidence also cuts against the theory that toughness alone reduces harm. The Sentencing Advisory Council's own data shows the age group now facing tougher bail and, from December 2025, adult sentencing exposure — children under 15 — already reoffend at the highest rate of any cohort in the system, 81% within eleven years of a Children's Court sentence, regardless of how tough or lenient that original sentence was. That does not prove tougher bail makes no difference in the short term; keeping a genuinely dangerous repeat offender off the street for a period does prevent further offending while detained. But it is evidence that punitive settings alone, without addressing what drives 10-to-14-year-olds back to offending, are unlikely to be the durable fix the "toughest bail laws in Australia" branding promises. And the remand data above shows the reversal's most immediate, measurable effect has fallen on Aboriginal children specifically — the cohort the Nelson inquest found was already most harmed by the system the 2023 reforms tried to fix, now recording detention numbers climbing back toward where they were before, with almost three-quarters of those detained not yet sentenced for anything.
The risk to the future
Another coronial finding, and a bail system nobody can plan around
Two risks compound each other. The first is the one the Nelson inquest already described: rising, unsentenced Aboriginal youth detention numbers are the precise pattern that produced a preventable death in an adult prison in 2020, and a Charter-incompatible sentencing law passed with only days of debate raises the likelihood that harm from the remand system will again land disproportionately on First Nations children and their families before anyone in government is required to answer for it. The second is a policy-design risk that outlasts any single law: a bail system that has moved from loosening to tightening to a Charter-breaching extreme in under three years gives police, courts, families and the children themselves no stable set of rules to plan around, and leaves the next government — of whichever party — inheriting a system that has not been allowed to run long enough, in either direction, to show whether its underlying theory actually works. Victoria will not get an honest answer to whether tougher bail reduces high-harm youth offending, or whether the Nelson reforms would have worked given time, because neither approach has yet been left in place long enough to be properly tested against the state's own data.