Summary of discussion

Transparency International New Zealand  |  23 September 2026  |  Hosted by MinterEllison  |  Chair: Sir Peter Boshier

The forum opened with a karakia and a welcome from Sir Peter Boshier, former Chief Ombudsman. Debbie Gee (Deputy Chair, TINZ) summarised the Green Paper, then led a panel of Prof Dean Knight (Victoria University of Wellington), Sia Aston (Department of Conservation), Richard Harman (journalist) and Sir Peter. Open discussion followed, with participants from media, academia, NGOs, government agencies and the Law Commission, in the room and online. Participants were told that remarks would not be attributed, so floor contributors are not named here.

1. Starting point: the Green Paper

The Green Paper is a “review of reviews”. It draws on the Law Commission (2012), the Chief Ombudsman (2015), Ministry of Justice engagement (2019) and recent cost modelling. Its core finding is that the basic design of the Act is sound: a presumption of release, case-by-case decisions and Ombudsman oversight. However, practice has drifted from what the law intended, and known gaps have never been closed. The question it poses is why, after nearly 15 years of agreement on the problems, the response has always been better practice and never a change to the law. Its seven issues were:

• Timeliness. Agencies treat day 20 as a target rather than an outer limit, and extensions have become routine. This was the biggest complaint in 2019. Complaints to the Ombudsman about delay can themselves take up to two years to resolve.
• Culture of avoidance. Close to 40% of 2019 submitters described the public service as risk-averse. There is documented discouragement of written records, and over 40% of staff surveyed did not know their chief executive’s attitude to disclosure.
• Ministerial interference. The 2015 investigation found cases of ministerial staff trying to narrow requests or change decisions. The protocol proposed then has not settled the concern.
• Weak enforcement. Ombudsman recommendations are effectively unenforceable and there is no penalty for ignoring the Act. By contrast, the UK can issue binding notices backed by contempt, and several Australian states have offences for destroying records.
• Coverage gaps. These include contracted-out services, inconsistent treatment of SOEs and parliamentary agencies sitting outside the Act. These are precisely the areas where transparency deters misconduct.
• Cost and sustainability. Request volumes have nearly quadrupled since 2016. The system costs about $180m a year, forecast to nearly double by 2030/31. Reducing demand should not mean making the right harder to exercise.
• Dated legislation. The Act is 44 years old and hard to use. The public interest test is buried, and Ombudsman guidance has become a necessary “translation”. By comparison, Canada mandates a review every five years, while New Zealand has gone 14 years without legislative change.

2. Key themes

A. Delay, triage and prioritisation

Delay was widely accepted as the central failure. Richard Harman said the 20-day limit makes the Act of little immediate use to media: “20 days is when you get the letter telling you they’re going for an extension.” Sia Aston gave DOC’s figures. Requests rose from 553 to 893 a year over a decade and are up another 40% this year. About two-thirds are complex and around a third need frontline staff. DOC meets the deadline 98.8% of the time, but 15% of requests still need extensions for consultation.

Sir Peter proposed statutory triage on a “medical model”: urgent, current matters come first (for example, information needed before submissions on a bill close), and low-value or grudge requests can wait. Dean Knight asked whether 20 days is still fit for purpose. He suggested a shorter track for targeted, urgent requests and more flexibility for large inbox-wide searches. Richard supported a fast track for urgent requests. Other suggestions included separating simple document requests from searches of emails and communications. An AI tool could guide requesters as they draft, or test the reasons given for urgency.

Differing views: An experienced investigative journalist argued that the Act is “requester-blind and information-blind”. On that view, no official should judge one requester or media outlet as more important than another. A participant representing public users of the Act said a two-speed system already exists that favours journalists and lawyers. They noted that an urgency mechanism already exists, and warned that triage could deepen the unfairness. Another participant said agencies can already prioritise if they choose. On the other side, a floor suggestion was to formally tier requests by requester type, favouring those who disseminate to the public. Sia saw the attraction but said it is unworkable in practice, because agencies often do not know who is asking. Sir Peter said he meant prioritising by urgency and relevance to the Act’s purposes, not the requester’s status. He did, however, also suggest that the legitimacy of both requester and request matters.

B. Scope, volume and repeat requesters

Sia described the burden of very broad and repeated requests. A media request about a previous OIA request took 10 staff and 50 hours of senior adviser time. DOC’s top five non-media requesters made 11% of all requests, and one person has made 370 requests over 10 years. Sir Peter and Sia both flagged AI-generated requests of enormous breadth. Sir Peter recommended clarifying the frivolous and vexatious refusal grounds, as the Law Commission proposed in 2012. Sia sought clearer tools for sustained, exceptional patterns of requests, with high thresholds and oversight, so that access “doesn’t depend on whether agencies find a person difficult”.

There was broad agreement on early, two-way engagement to narrow scope. Agencies should contact requesters promptly and ask what they are really looking for. Sir Peter wanted room for informal, oral requests and answers. In response to an online question, he noted that Māori participants had preferred to sit down and talk a request through rather than receive a bureaucratic written answer. He suggested this could be built into a new Act.

Differing views: Some participants rejected “glib” talk of vexatious requesters. They pointed out that high-volume requesters are often citizen watchdogs, for example someone tracking police statistics that are not published. Measures aimed at requesters could also require identity checks, undermining the Green Paper’s proposal to remove the eligibility rules. Those rules are already used to delay requests. Sir Peter replied that the vexatious test attaches to the request, not the person, and that a filter for, say, the tenth request on the same issue is workable. Another participant said bad faith also runs the other way, with agencies engaging in “malicious compliance”. Examples were clarification questions sent on day 20 and large data dumps that bury what was sought.

C. Culture, leadership and political influence

Everyone agreed culture was central, but views differed on its source. Richard said culture is set by ministers: an evasive minister produces an evasive agency. He cited the no surprises policy, political advisers, sanitised records that omit real policy debate, and historic tactics such as unnumbered Friday-afternoon document dumps. Dean suggested rewriting the no surprises principle in the Cabinet Manual because it creates perverse incentives. He also asked why OIA decisions are treated as a mechanical withholding exercise. In his view, the Act’s purposes should shape them in the same way as any other statutory decision. He noted that requests are often processed by junior staff, and floated transparency champions within agencies.

Sia said most OIA staff are earnest and detail-focused. Delay often reflects over-complication and genuine legal risk rather than obstruction. She also urged agencies to accept that sensitive material will come out and prepare to explain it. Sir Peter said years of promoting good leadership had not been enough, and that a “circuit breaker” was needed. A recent Ombudsman review of five agencies had found risk aversion put ahead of timely release at one of them.

D. Accountability, enforcement and sanctions

Sir Peter called the lack of consequences “lamentable”: “a telling off by the Ombudsman is about as good as it gets.” He sought discussion on indirect sanctions, such as making chief executives accountable through performance measures, rather than punishing junior staff. Richard noted a suggestion in the previous day’s parliamentary debate to apply the Crimes Act offence of obstructing a statute. Dean framed the choice as between two approaches, both relevant: sanctions for misuse, or continuous improvement through a trust-based oversight relationship. He also asked whether the information function should move from the Ombudsman to a dedicated body within a wider “integrity branch” of government.

Differing views: A leading OIA user argued that sanctions would punish the wrong people, namely staff acting under spoken or unspoken pressure from above. Sia warned they could produce even more “vanilla” records and more reluctance to write things down, though she thought the idea was worth exploring. Others pressed for real consequences for agencies acting in bad faith.

E. Proactive release

Dean named proactive release as the single reform he would back. It is transparency in itself, and it shows officials and ministers that “the sky doesn’t fall”, as the pandemic demonstrated. He said it needs a proper statutory footing that fits with the request system. Sia agreed it is important (DOC proactively releases about 24% of responses) but said it is not free and lacks the legal protection that OIA releases have. Sir Peter recommended extending the good-faith protection to proactive release. Examples of good practice were police’s monthly buyback data after the Christchurch attacks, and an international infrastructure model. That model publishes a defined set of around a dozen project and contract data points, with stakeholder engagement and reactive requests as further layers.

Differing views: Some participants reported proactive release being used to stall requests (“it will be published soon”). They also raised submissions being due before Cabinet papers are released, and document stamps and formatting that make PDFs unsearchable.

F. Information management, records and technology

A frequent requester argued that “the problem is not the OIA; the problem is the government”. Decades of under-investment have left information systems unable to keep pace with the Act. He gave the example of five years of NZDF records in unmarked boxes and email periods described as unsearchable. Sia agreed that systems are a “hodgepodge” across email, Teams, texts, databases and paper, and that fixing them is a matter for agencies rather than the Act. Several participants said the Public Records Act is integral: failing to create or keep records defeats accountability. They also raised enforcement (no prosecutions to date) and reduced funding for Archives.

Differing views: On AI, Sir Peter saw strong potential for triage and for decision support, given how predictable much Ombudsman precedent is. Sia said DOC uses it to classify about 30% of requests, but that results on even simple requests have been poor and human judgement remains essential. A requester argued that AI would only help once it was built on sound record-keeping.

G. Legislative design, scope and purpose

Dean and Sir Peter saw the statutory scheme as fundamentally sound, needing modernisation rather than replacement. They located the main deficiency in practice. Debbie noted that the Green Paper nonetheless asks whether a full rewrite is due.

Differing views: One participant rejected the claim that the Act is fundamentally sound and said the discussion had become too “managerial”. They framed access to information as a human right under the ICCPR, and noted that New Zealand now falls short of international standards. In particular, there is no public interest override for the section 6 withholding grounds. They also argued that decades of focus on the accountability purpose, which bureaucracies are skilled at gaming, has produced an adversarial culture at the expense of participation. Richard and Sia doubted that the OIA is the main vehicle for participation, given the other channels available. Sir Peter said requests should be focused on the participation purpose.

On LGOIMA, Sir Peter saw merit in bringing local and central government under one Act, noting that small councils struggle with capacity. Others observed that local government tends to be more forthcoming because it has less of a political layer. One participant noted that LGOIMA has no eligibility rules and has “worked fine”. Universities were described as an awkward fit.

H. The wider transparency ecosystem

Richard and Dean both placed the OIA within a wider system that is weakening in places. Richard pointed to the Privacy Act limiting release, the planned loss of regulatory impact statements, written parliamentary questions used as “drift-netting”, and Scrutiny Week reduced to grandstanding. Dean cautioned against routing all journalist–official contact through the formal OIA process. He pointed to other forms of accountability such as post-Cabinet press conferences. Sia cited community engagement, about 1,200 media enquiries a year at DOC, and ministers who are unusually accessible by international standards.

3. The one thing that would make a difference

Sir Peter called for a “circuit breaker”: a single, concrete change that would shift the system. A floor participant sharpened this before the break. They asked for one reform that would be politically feasible, have real impact, be simple to draft and be easy for the public to understand. Otherwise, they warned, “we’re just going to be here in five years’ time”. They noted that parties in opposition promise a review and then defer it in government. They also cautioned that much of the problem is cultural and cannot be fixed only by changing the statute book. The panel was asked to answer after the break. Their answers were:

• Dean Knight: proactive disclosure. It reflects transparency as a principle in its own right.  A stronger requirement and more use of proactive disclosure would shift defensive culture over time as officials and ministers see that openness does not cause harm.
• Sir Peter: statutory triage. He would define the scope of a request more tightly and prioritise responses by urgency and relevance to the Act’s purposes, set out in the Act itself.  He would also add good-faith protection for proactive release as “one of the ingredients in the cake”.
• Sia Aston: tools for repeat and very broad requests. She wants a way for agencies to push back on these, including AI-generated requests, without undermining the Act’s purpose, so their time goes to high-value requests. One participant replied that the Act already allows such requests to be refused.
• Richard Harman: a fast track for urgent requests. He backed Sir Peter’s triage idea, but said the underlying problem is culture: “fast track, but just keep working on the culture”.
• From the floor: a better proactive release regime. One participant would put most effort here too, but noted that the promise of proactive release is already used to stall requests, and that the expected one-month timeframe is not observed.

Proactive release and some form of prioritisation came up most often. Both carried caveats raised elsewhere in the discussion: proactive release must not become a reason to delay requests, and triage must not deepen the existing two-speed system (see themes A and E).

4. Suggestions for improvement

To the Act

• Statutory triage: a faster track for urgent or time-critical requests and more flexibility for very large searches. The trade-offs for requester neutrality were contested (see A).
• Review whether the 20-day limit fits today’s volume of digital records, and modernise the Act’s language and concepts for email, messaging, databases and AI-assisted search.
• Clearer frivolous and vexatious provisions, and tools for sustained patterns of requests, with high thresholds and oversight (contested).
• Stronger enforcement: binding Ombudsman or regulator powers, accountability at chief executive level, and possible offences for wilful obstruction or record destruction (contested).
• A statutory footing for proactive release, including good-faith protection equivalent to that for OIA releases.
• Add a public interest test to the section 6 grounds, remove eligibility rules, and close gaps in coverage (contracted services, SOEs, parliamentary agencies).
• Consider a separate information regulator, merging OIA and LGOIMA, a mandatory periodic review (the Canadian model), or a full rewrite.
• Provide for oral or conversational engagement, including approaches that reflect tikanga Māori.

To practice and the wider ecosystem

• Rewrite the no surprises principle in the Cabinet Manual and revisit the ministerial consultation protocol.
• Invest in information management systems. Strengthen the Public Records Act and its enforcement, and fund Archives.
• Get back to requesters early to refine scope, give plain-English help, and use AI-assisted drafting tools that also test urgency claims.
• Give OIA work senior leadership and transparency champions, and make sure processing staff have enough capability.
• Make proactive release timely, searchable and free of obstructive stamps, release it ahead of consultation deadlines, and do not use it to delay requests.
• Consider funding the OIA function through Parliament rather than departmental baselines.
• Protect other transparency channels, such as regulatory impact statements, written questions and select committee scrutiny.

5. Next steps

Sir Peter thanked participants and panellists, and thanked Debbie for the Green Paper. He said the action is needed. The proactively released paper from the Ministry of Justice indicates that reforms are possible, so there is a platform to influence. TINZ will produce this summary and then a white paper with recommendations.  Participants were invited to make submissions.

‍