Goldwater v. Phoenix: No Rigid “Probability of Harm” Element Under Carlson; Appellate De Novo Rebalancing and In Camera Review for Negotiation-Stage Public Records

1) Introduction

In GOLDWATER v. PHOENIX (Ariz. July 17, 2026), the Arizona Supreme Court clarified how courts apply the common-law “best interests of the state” limitation on disclosure of public records under A.R.S. § 39-121, commonly analyzed through the Carlson balancing test.

The Barry Goldwater Institute for Public Policy Research Center (“Goldwater”) sought public records from the City of Phoenix (“City”) concerning ongoing collective bargaining negotiations with the Phoenix Law Enforcement Association (“PLEA”), including draft Memoranda of Understanding (“MOU”) and bargaining proposals. The City refused to disclose certain negotiation-stage materials, invoking the “best interests of the state” on the theory that disclosure would chill candid negotiations and cause harmful politicization and strategic behavior.

The case presented two recurring issues: (1) whether the “best interests” exception requires a public entity to prove it is more likely than not that disclosure will cause “specific, material harm,” and (2) how appellate courts should review trial-court decisions applying Carlson. The Supreme Court also addressed the role of in camera review and the permissibility of delayed disclosure tied to the negotiation process.

2) Summary of the Opinion

  • No rigid “more-likely-than-not” element: The Court held that a public entity invoking the “best interests of the state” is not required to prove, as a discrete threshold element, that disclosure is probable (greater than 50%) to cause “specific, material harm” (discussing the phrasing in Mitchell v. Superior Court).
  • But harm must be specific, material, and non-speculative: The entity still bears the burden under Carlson to show (a) a specific, material harm, (b) some meaningful degree of likelihood of that harm, and (c) a causal connection between disclosure and the harm; vague or purely speculative assertions do not suffice.
  • Appellate standard of review clarified: Appellate courts defer to trial-court fact findings unless clearly erroneous, but must review de novo the legal conclusion—i.e., must independently determine whether the evidence shows the “best interests” outweigh the presumption of disclosure.
  • In camera review required on remand: The Court ordered remand for in camera review (with redacted and unredacted versions) of responsive records across categories not already disclosed, to allow tailored disclosure/redaction.
  • Delayed disclosure: If the “next MOU” has already been finalized, the City must disclose records the trial court ordered released and cannot wait for the litigation to end.

The Supreme Court reversed and remanded and vacated the court of appeals’ opinion.

3) Analysis

A. Precedents Cited

The Court’s reasoning is built on a line of Arizona public-records cases defining (1) the presumption of disclosure, (2) the nature of countervailing interests, and (3) the procedural tools courts should use to separate disclosable information from legitimately withheld material.

Carlson v. Pima County, 141 Ariz. 487 (1984)

Carlson is the foundational case establishing that although Arizona public-records law strongly favors access, officials may deny or restrict access where “the interests of privacy, confidentiality, or the best interest of the state in carrying out its legitimate activities outweigh the general policy of open access.” The Court in GOLDWATER v. PHOENIX stressed that Carlson is a balancing test, not a rigid checklist of elements. That framing underpins the Court’s rejection of a categorical “probability > 50%” prerequisite as inconsistent with true balancing.

Mitchell v. Superior Court, 142 Ariz. 332 (1984)

Mitchell contains the oft-quoted statement that the non-disclosing party bears the burden of showing “the probability that specific, material harm will result from disclosure.” Goldwater argued this language creates a strict “more likely than not” requirement. The Supreme Court disagreed, reading Mitchell as reinforcing the burden allocation within balancing—not imposing a universal threshold element that must be proven before balancing can occur. The Court preserved the Mitchell concept as a likelihood factor within balancing.

Arizona Bd. of Regents v. Phoenix Newspapers, Inc., 167 Ariz. 254 (1991)

This case upheld nondisclosure of names of university presidential “prospects” because disclosure “could chill” recruitment and harm the search process. GOLDWATER v. PHOENIX used it to demonstrate that Arizona courts have allowed withholding based on demonstrated, experience-based risks to governmental functions without imposing a rigid “more likely than not” element in every case.

Cox Ariz. Publ'ns, Inc. v. Collins, 175 Ariz. 11 (1993)

Cox supplies the counterpoint: generalized claims and “global generalities” about possible harms from disclosure of police records were insufficient. The Supreme Court relied on Cox to reinforce that although strict probability is not a threshold element, the harm must still be articulated with specificity and grounded in evidence—not speculation.

Scottsdale Unified School Dist. No. 48 of Maricopa County v. KPNX Broadcasting Co., 191 Ariz. 297 (1998)

This case illustrates application of balancing (teacher birth dates) and clarifies that appellate courts defer to fact findings but may draw their own legal conclusions. GOLDWATER v. PHOENIX invoked it for the de novo review principle in public-records disputes.

Griffis v. Pinal County, 215 Ariz. 1 (2007)

Cited for the proposition that Arizona public-records law creates a “strong presumption in favor of disclosure” once a document qualifies as a public record—a presumption the City had to overcome under Carlson.

Phoenix Newspapers, Inc. v. Keegan, 201 Ariz. 344 (App. 2001)

Used to clarify that “best interests of the state” is not the narrow interest of the custodian or agency alone; it includes the overall interests of government and the people, including impacts on the agency’s mission and the public. The Supreme Court adopted that broader understanding, while cautioning that municipalities need not prove harm to the entire state in every case.

Hodai v. City of Tucson, 239 Ariz. 34 (App. 2016)

Hodai exemplifies “specific, material harm” in the context of sensitive investigative techniques and illustrates that “vague assertions” are insufficient. The Supreme Court used it to show how specificity and a tight causal link support withholding, without converting the inquiry into a rigid probability element.

Smith v. Town of Marana, 254 Ariz. 393 (App. 2022)

Cited for the proposition that speculation and generalized potential harms do not satisfy the Carlson/Mitchell standard. The Supreme Court used Smith to set a floor: the likelihood of harm cannot be “de minimis, vague, or purely speculative.”

Mathews v. Pyle, 75 Ariz. 76 (1952) and Little v. Gilkinson, 130 Ariz. 415 (App. 1981)

These cases were relied upon to support in camera review as a principal method to resolve disputes about confidentiality and “best interests” claims. The Court quoted Mathews to underscore the necessity and utility of in camera inspection in deciding whether disclosure would be “detrimental to the best interests of the state,” and it highlighted Carlson’s emphasis on redaction as an alternative to total withholding.

Additional authorities

  • Cheatham v. DiCiccio, 240 Ariz. 314 (2016): referenced to contextualize MOUs and topics they govern in Phoenix’s meet-and-confer regime.
  • Dep't of Air Force v. Rose, 425 U.S. 352 (1976), and U.S. Dep't of Just. v. Reps. Comm. for Freedom of Press, 489 U.S. 749 (1989): cited for the transparency rationale—opening “agency action to the light of public scrutiny” and enabling citizens “to be informed about what their government is up to.”
  • Conard v. Dillingham, 23 Ariz. 596 (1922): cited for a definition of “probable” supporting Goldwater’s argument, which the Court ultimately rejected as a rigid requirement.
  • Arizona Republican Party v. Richer, 257 Ariz. 237 (2024): cited for the “clearly erroneous” standard on factual findings.
  • Hurd v. Hurd, 223 Ariz. 48 (App. 2009), and In re Estate of Pouser, 193 Ariz. 574 (1999): cited by the court of appeals, but criticized by the Supreme Court as inapposite to public-records appellate review.

B. Legal Reasoning

1. Reframing “probability” as a factor within balancing—not a mandatory element

The Court’s central doctrinal move is interpretive: it reconciles Mitchell’s “probability” phrasing with Carlson’s structure as a true balancing test. Under Carlson, courts weigh the strong presumption of openness against countervailing interests (privacy, confidentiality, and the best interests of the state).

The Court held that importing a strict “more likely than not” requirement would distort balancing in at least two ways: (1) it could force disclosure until harm is already demonstrated (or until government can prove a >50% likelihood in advance), and (2) it fails to account for scenarios where the likelihood of harm is uncertain but the magnitude could be catastrophic. The result: “probability” is not a gatekeeping element; rather, the degree of likelihood is one of multiple variables that determine whether the “best interests” outweigh disclosure.

2. The burden remains meaningful: non-disclosure must be supported by evidence of specific, material harm

The Court reinforced that relaxing “probability” as a rigid element does not permit agencies to withhold records by incantation. To prevent “best interests” from swallowing the rule of access, the Court emphasized:

  • Specificity: the harm must be articulated concretely, not in generalized terms;
  • Materiality: the harm must be significant enough to justify overriding the statutory presumption;
  • Likelihood: the risk cannot be de minimis or purely speculative;
  • Causation: there must be a credible causal connection between disclosure and the asserted harm;
  • Narrow tailoring/temporality: if the harm is time-bound (e.g., tied to ongoing negotiations), a delay in disclosure should be temporary and limited to that harm’s duration.

3. “Best interests of the state” for municipalities: the public’s interest, not statewide harm in every case

A notable clarification concerns terminology: in municipal cases, “state” does not require proof that the entire State of Arizona will be harmed. The Court adopted the understanding reflected in Phoenix Newspapers, Inc. v. Keegan that “best interests” includes the broader interests of government and the people, including effects on agency mission and how the public would be affected— but without imposing a statewide-harm showing as a categorical requirement for local entities.

4. Appellate review: defer on facts, re-balance de novo

The Court corrected what it viewed as doctrinal drift: public-records disputes involve factfinding (e.g., what harms are credibly supported) and a legal conclusion (whether the supported interests outweigh openness).

  • Findings of fact: reviewed deferentially (clearly erroneous standard).
  • Balancing conclusion: reviewed de novo, meaning the appellate court must independently decide whether the record meets the Carlson standard; it cannot simply “decline to reweigh” in a way that abdicates the legal judgment Carlson requires.

5. In camera review and redaction as core tools

The Court emphasized that in camera review is often essential where “best interests” is invoked, relying on Mathews v. Pyle, Carlson v. Pima County, and Mitchell v. Superior Court. It ordered the City to provide both unredacted records and redacted versions reflecting its claimed bases for withholding.

This approach operationalizes two transparency-protective principles: (1) courts should verify claims of harm against the actual contents, and (2) where possible, courts should prefer targeted redactions over categorical withholding.

C. Impact

1. Practical guidance for agencies resisting disclosure

Public entities now have clearer instructions: they need not prove a >50% likelihood of harm as a standalone element, but they must present evidence that identifies a specific, material harm, explains why disclosure would contribute to it, and shows a meaningful likelihood—more than vague predictions. Agencies should anticipate that courts will demand a record suitable for in camera scrutiny and potential redaction.

2. Litigation and appellate posture in public-records cases

The opinion strengthens appellate oversight. By requiring de novo review of the balancing conclusion, the Court promotes statewide uniformity in the application of Carlson, reducing the risk that disclosure will hinge largely on trial-level discretion unchecked on appeal.

3. Negotiation-stage records and “delayed disclosure” frameworks

For collective bargaining and similar negotiation contexts, the Court did not announce a blanket exemption for drafts or proposals. Instead, it insisted on document-specific evaluation through in camera review, attentive to: (a) whether claimed harms are tied to ongoing negotiations, (b) whether some provisions are boilerplate already public, and (c) whether any delay in disclosure ends when the harm ends (e.g., after the “next MOU” is finalized).

4. Transparency policy: the presumption remains robust

The opinion preserves the “strong presumption” of disclosure and rejects both extremes: (1) a rigid probability threshold that could undervalue low-likelihood/high-severity harms, and (2) a permissive “best interests” invocation that would enable secrecy by generalized assertion.

4) Complex Concepts Simplified

  • Public record (A.R.S. § 39-121): Government-held documents generally open for inspection during office hours.
  • Carlson balancing test: A court weighs the presumption of openness against countervailing interests (privacy, confidentiality, or the government’s ability to carry out legitimate activities).
  • “Specific, material harm”: A concrete, significant harm—not a vague “it could be bad.”
  • Likelihood (as used here): Not a strict “more likely than not” element, but the risk cannot be trivial or purely speculative; the court weighs how likely and how severe the harm is.
  • Causal connection: The agency must show how disclosure would lead to the harm, not just that harm could exist generally.
  • In camera review: The judge privately reviews documents to decide what can be disclosed or redacted.
  • Redaction: Blacking out limited protected portions while releasing the remainder.
  • Standards of review: “Clearly erroneous” means appellate courts usually accept trial-level fact findings unless plainly wrong; “de novo” means the appellate court decides the legal question independently.

5) Conclusion

GOLDWATER v. PHOENIX modernizes and clarifies Arizona’s common-law public-records doctrine in three key ways: (1) it rejects a rigid, standalone “more likely than not” probability element while preserving a meaningful, evidence-based burden to show specific, material, non-speculative harm and causation; (2) it requires appellate courts to conduct de novo review of the balancing conclusion, ensuring uniform application of Carlson; and (3) it elevates in camera review and redaction as essential mechanisms for tailoring disclosure, especially in negotiation-stage disputes.

The decision thus reinforces transparency as the default while providing a structured, workable method to protect legitimate governmental interests without allowing “best interests of the state” to become an all-purpose shield against disclosure.