Oversight & Accountability

A Named Absence: Georgia's Missing Oversight Body

Georgia is not a state with weak prison oversight. It is a state with none of the independent kind. The National Resource Center for Correctional Oversight's interactive map counts 19 states plus the District of Columbia with an external, independent prison oversight body; a January 2025 count by Deitch puts the figure at "at least 20 states and Washington, D.C." (Nobody Watches the Watchmen: Independent Correctional Oversight Models and the Statutory Path for Georgia). Georgia appears on neither list, and the Brennan Center's 2026 comparison report names the state among those lacking independent oversight mechanisms (Comparative Solutions Evidence Base; Nobody Watches the Watchmen). Georgia has no independent correctional ombudsman, inspector general, oversight commission, or authorized nonprofit with access to its prisons. All PREA monitoring is conducted internally by GDC's Office of Professional Standards, which reports to the GDC Commissioner; the internal Ombudsman Unit is not independent — it sits inside OPS within GDC (Sexual Violence & PREA Compliance in Georgia Prisons).

This is not an untested idea that simply never came up. The National Prison Rape Elimination Commission's 2009 report stated plainly that "dramatic reductions in sexual abuse depend on rigorous internal monitoring and external oversight" (Sexual Violence & PREA Compliance in Georgia Prisons). In December 2024 the Georgia Senate Study Committee on the Department of Corrections heard testimony on the Federal Prison Oversight Act of 2024 — independent oversight through an ombudsman plus regular Inspector General inspections, with corrective action required within 60 days — noting that Indiana, New Jersey, Pennsylvania, and Virginia already run versions of that model (2024 Georgia Senate Study Committee on the Department of Corrections — Final Report (SR 570)). Congress passed the federal version in July 2024, championed by Senator Jon Ossoff, mandating DOJ IG inspections of all 122 federal prisons and creating an independent ombudsman (Sexual Violence & PREA Compliance in Georgia Prisons). At the same committee, a separate proposal by Sen. Bearden to have the GBI investigate all in-custody deaths also did not make the final list (Nobody Watches the Watchmen).

Georgia has been under external prison oversight before. Guthrie v. Evans documents a federal court takeover of Georgia State Prison running from 1972 to 1999, after the court found conditions unconstitutional (Guthrie v. Evans: The Federal Court Takeover of Georgia State Prison (1972-1999)). GPS's reading of that record — an analytic through-line rather than a court holding — is that federal supervision imposed standards, the PLRA terminated it, and the state then moved to circumvent those standards, with the 2024 DOJ findings re-identifying the same categories of constitutional violation fifty years later (#4766). The state also knows how to exit oversight: the Department of Juvenile Justice completed all requirements of its DOJ memorandum of agreement in May 2009, ending eleven years of federal supervision (BOPP Annual Report FY2012 — Georgia Parole Board).

Even the internal paper trail is incomplete. Of the 141 months from January 2015 to September 2026, 31 hold neither a minutes document nor a summary in the Georgia Board of Corrections meeting record (Georgia Board of Corrections meeting record 2015-2026 corpus). A board that meets regularly but leaves a third of its recent record undocumented is not a substitute for a body with a statutory mandate to inspect, compel, and publish.

Death Without a Trace: A System Built Not to Know

Georgia's death-investigation architecture disperses authority to people without forensic training and then allows the answer to remain permanently open. In 155 of the state's 159 counties, the official who takes charge of a body and can sign a death certificate is an elected coroner (Who Decides How They Died). The statutory qualifications are: at least 25 years old, a U.S. citizen, a registered voter, a high school diploma, and no felony conviction — no medical training required, with a 40-hour course and 24 hours of annual in-service training standing in for a medical education (#7167). Only DeKalb, Fulton, Cobb, and Gwinnett have abolished the office in favor of a county medical examiner (#7169). Under O.C.G.A. § 45-16-22, the medical examiner has sole discretion to determine whether an autopsy is needed (#8140); a coroner may certify a death without one, and cause of death may be entered as "pending" and never amended (#8141). The pending-investigation exemption under the Open Records Act is the primary legal mechanism by which cause-of-death information stays unavailable (#8150).

GDC controls the front end of that chain. Under SOP 208.03, it is the GDC Criminal Investigations Division Inspector who decides whether to notify the GBI about a death — meaning the agency gatekeeps outside involvement in its own deaths (#8135). The October 2024 DOJ findings documented that GDC obstructed federal investigators, conducted pre-inspection cleanups, and refused to produce documents until a subpoena was enforced (#8136, #8151), and concluded that GDC's deficient reporting and investigation practices prevent adequate response to a substantial risk of harm (#8167).

The output of that system is a record that cannot be reconciled with itself. DOJ found that GDC "inaccurately reports these deaths both internally and externally, and in a manner that underreports the extent of violence and homicide" (#5006). In June 2024, GDC reported 6 killings; DOJ found at least 18 (#3183). DOJ specifically found that GDC categorized obvious homicides as "unknown" causes (#5943). Two decades before that, a Georgia Department of Audits and Accounts review found that only 30 of 233 inmate deaths in 2005-06 (excluding executions) had evidence of a mortality review, and the Office of Health Services could not locate medical files for 86 of the 203 unreviewed deaths (#13749, #13750). A 2010 state audit found the State Medical Examiner's Office was not contacted in 26 percent of statutorily required suicide referrals (#8139). Nationally, the GAO found in 2022 that nearly 1,000 deaths that potentially should have been reported to DOJ under the Death in Custody Reporting Act were not, and that 70 percent of state-provided records were missing at least one required element; BJS stopped publishing detailed prison mortality data after reference year 2019 (#8121). A Senate subcommittee concluded in September 2022 that DOJ's failure to implement the DCRA deprived Congress and the public of information about who is dying in custody and why (#8154).

GPS's own tracking illustrates how far the published record sits from any verified total. GPS mortality records — floor counts, because GDC reports deaths roughly two months late, and beginning only in 2020 — contain 1,890 deaths since January 2020: 704 Unknown/Pending, 664 Natural/Medical, 263 Homicide, 118 Suicide, 58 Overdose, and smaller categories. GPS separately maintains a homicide database of 244 confirmed homicides, 170 of which were reclassified using AJC investigation data revealing systematic misclassification (#5502). DOJ, using a different window, documented 142 homicides from 2018 through 2023 (#7307). These counts use different periods and definitions, and no state agency publishes a reconciled total. That unreconciled spread is itself the accountability finding — and it has widened by design: in March 2024, GDC's Commissioner stopped including preliminary cause of death in monthly mortality reports, and the department routinely blacks out entire pages of incident reports (#136, #164).

The Federal Finding and the Enforcement Question

The most detailed account of Georgia's prison system in state history was produced by the federal government, and it has not been enforced. The DOJ Civil Rights Division opened a CRIPA pattern-or-practice investigation into GDC in February 2016, initially focused on protection of LGBTI people from sexual abuse (#6534, #6011), expanded it in September 2021 to cover violence against all medium- and close-security prisoners (#6535), and expanded again in April 2024 to add restrictive housing, disciplinary practices, and special education services (#6661). Investigators visited 17 of roughly 34 state prisons in 2022-2023 (#1612, #2529), conducted hundreds of private one-on-one interviews and reviewed tens of thousands of records (#1613, #7123), and received more than 1,000 letters and communications from incarcerated people and their families (#2532). GDC set every site visit weeks or months in advance, required Special Operations staff to facilitate, refused to let the DOJ group split up, and did not permit spontaneous touring (#2587).

On October 1, 2024, DOJ issued a 93-page findings report concluding that Georgia "engages in a pattern or practice of violating incarcerated persons' constitutional rights" under the Eighth Amendment, and that the State is deliberately indifferent (#8084, #7231). It made two formal findings: failure to protect from violence by other incarcerated people, and failure to protect from sexual violence, including specific findings on LGBTI prisoners (#6672, #6673). It documented 142 homicides from 2018 through 2023 — 48 in the first three years, 94 in the latter three, a 95.8 percent increase (#7307). It recommended 82 remedial measures (#8088), spanning 12 pages, including filling at least 90 percent of allocated correctional officer posts, documented and investigated violent-incident response, re-evaluation of housing and classification, weekly contraband searches, and an overhaul of sexual-abuse prevention (#6659). Separately, and bearing on the Tier system without being a finding about it, the same report found that "GDC fails to control violence even in its segregated housing units and exposes incarcerated persons to an unreasonable risk of harm due to its inappropriate use of segregated housing" — a finding made under a violence-and-safety scope covering medium- and close-security facilities, not a finding about the Tier programme as such (#19311). GDC rejected the findings the same day (#3902, #8090, #5762, #6660).

Enforcement has not followed. The 49-day deadline DOJ set for Georgia to respond passed without action (#3878). In January 2025, GDC confirmed DOJ had sent a settlement proposal under review (#6043). As of April 2026, no consent decree had been reached (#5832); as of May 18, 2026, DOJ had not filed a CRIPA enforcement action (#7349). No new federal §242 prosecution of GDC staff has followed the findings (#6704) — and in the entire FY2018-forward window, research produced exactly one published §241/§242 case against GDC sworn staff (#6703). The Trump administration's DOJ has moved to dismiss consent decrees and halt reform investigations nationally (#6042) and has closed pattern-or-practice investigations into multiple police departments (#1630). The April 2024 restrictive-housing prong of the Georgia investigation has still not produced public findings (#6661). That prong is the live external route to a Tier population count: a CRIPA findings report on restrictive housing would ordinarily contain population, duration and per-facility figures obtained by compulsory process, and it remains unissued (#19311). Until it is, the absence of those findings is a dated gap and not evidence about Georgia's Tier population either way (#19311). As one GPS collection puts it, federal involvement in state prison accountability is episodic, slow, and politically contingent (#6559).

One finding in the same report points the other direction. DOJ found that Walker State Prison, with better staffing and more robust programming, had no reported homicides in several years and less violence and contraband than the other 16 prisons visited — and stated that Walker "shows that larger-scale improvement is possible with an appropriate strategy and sufficient resources" (#13688, #13689).

Internal Filters: Grievances, OPS, and the Discipline Gap

Before a Georgia prisoner can file a federal civil-rights claim, they must exhaust GDC's internal grievance process — a process that can close the door before the law is reached. A grievance alleging staff use of force that is deemed "Non-Compliance" is forwarded to the Office of Professional Standards and the Criminal Investigation Division, and if accepted, the decision is expressly non-appealable — cutting off the appeal step that Woodford-style proper exhaustion would otherwise require (#8022). In Varner v. Shepard, 852 F. App'x 411 (11th Cir. 2021), a §1983 claim against Augusta State Medical Prison officials was dismissed with prejudice because the prisoner filed no timely grievance and GDC did not waive the procedural defects (#7953). GPS's case-management aggregation describes grievances that were "lost" en route, never returned, or pressured into informal resolution that left no paper trail (#6582). The audits that would test any of this are required but unpublished: SOP 227.02 requires OPS Audits and Compliance to audit the grievance process at least annually, and to generate semi-annual reports to the Commissioner on the number and type of grievances filed — neither is published (#7943, #8035, #8034).

The same procedure prohibits retaliation in categorical terms and then routes the alleged act out of the system that would test it. SOP 227.02, "Statewide Grievance Procedure," effective 2019-05-10, provides in Section IV.A.3 that "Retaliation against an Offender for filing a grievance is strictly prohibited," and that "The prohibited retaliation includes, but is not limited to, disciplinary action against the Offender for filing a grievance"; Section IV.B.3 adds that "An Offender may file a grievance alleging retaliation or harassment, regardless of the form of the alleged retaliation or harassment" (#19319). But Section IV.B.2.d lists as non-grievable "Involuntary assignments to Administrative Segregation," directing the appeal to SOP 209.06, and Section IV.B.2.g lists as non-grievable "Housing assignments, program assignments, security classifications or work assignments, unless there is an alleged threat to the Offender's health or safety" (#19319). A person moved into administrative segregation after filing a grievance may therefore file a retaliation grievance in the abstract but may not grieve the segregation assignment itself unless able to frame a threat to health or safety. The remedy is routed into the SOP 209.06 classification appeal, which tests whether the placement criteria are met and does not test motive, and which terminates with the Warden, whose "review and decision on the appeal is final" (#19319).

The one record that would show a grievance preceded a Tier placement is, by GDC's own rule, missing from the file that decides release. SOP 227.02 provides that "A Grievance must not be placed in the Offender's file or referred to in SCRIBE case notes" (#19320), while SOP 209.08 conditions release from Tier II on the Unit Manager's review in SCRIBE of the offender's disciplinary reports, profiles, case notes, Security Threat Group history, incident reports, progress notes, movement history, mental-health status, personal history sheet, medical profile, court production orders, and county of conviction and residence (#19320). A keyword search of SOP 209.08 for "grievance" returns nothing: the Tier II policy does not mention grievances at all and contains no anti-retaliation provision (#19320).

Where the internal investigative machinery does run, it runs quietly. According to GDC's own incident records from 22 prisons between January 2022 and April 2023, less than 10 percent of fights were forwarded to OPS for investigation; less than 23 percent of inmate-on-inmate assaults; less than 12 percent of incidents involving serious injury; and less than 6 percent of weapon incidents (#2610, #2611, #2612, #2613). GDC informed DOJ that there are no centralized policies or procedures governing facility-level investigations or incident reviews (#2629), and the OPS director cited a prior "communication breakdown" between OPS and the Facilities Division, with wardens reporting they do not receive investigation reports or summaries (#2628). The consequence is not hypothetical: a March 2020 assault at Coastal State Prison was forwarded to OPS with no records indicating any investigation took place, and the same assailant re-entered the system in 2022 and strangled his cellmate to death at GDCP (#2627). In the Mitchell case, GDC's own review was described as "superficial" and produced retraining recommendations rather than termination or prosecution; the supervisor who prepared a false incident report was not prosecuted (#6775).

The scale of the staff misconduct problem is documented even where the consequences are not. At least 428 GDC employees were arrested for on-the-job criminal conduct between January 2018 and September 2023 — an average of more than seven per month (#6681), with roughly 80 percent involving contraband (#6808). Operation Ghost Guard (FBI/GDC, 2014-2016) indicted approximately 130 subjects, 47 of them correctional officers, and found "criminal and corrupt activities" in 11 of the state's 35 corrections facilities — nearly one-third of all GDC prisons (#6732, #6734). Operation Skyhawk (concluded March 2024) produced 150 arrests including eight correctional officers, more than 1,000 charges, and $7 million in seized contraband (#6721). Individual cases ran to the top of facility command: Smith State Prison Warden Brian Dennis Adams was arrested by GBI on February 8, 2023 on RICO, bribery, false statements, and violation of oath charges connected to the YSL Squad smuggling and murder ring, and terminated the same day (#6713, #6806). In August 2024, a GDC Office of Professional Standards investigator was among five people charged in the Floyd County Jail beating — direct evidence that the agency's internal investigators have at least one publicly named compromise (#6731). The internal architecture that is supposed to catch these cases is not public: GDC internal investigation reports are classified as confidential state secrets unless the Commissioner declassifies them in writing (#8079).

Indefinite Segregation: The Tier System and Its Unpublished Count

Georgia's segregation architecture is tiered, multi-site, and — on the state's own policy record — has no published population count. GDC operates four program layers: Tier I (short-term disciplinary, protective-custody, and transient housing), Tier II (long-term administrative segregation), Tier III (the Special Management Unit at Georgia Diagnostic and Classification Prison), and STEP step-down programs attached to Tier II and Tier III. A separate policy, SOP 209.11, governs restrictive housing for juveniles under 18. Tier I, under SOP 209.07 (effective 04/30/2015), is not limited to close-security prisons: its applicability section names "All State Prisons, County CIs, Private Prisons, Intensive Treatment Facilities, Transitional Centers, Probation Detention Centers, and Probation Boot Camp housing GDC offenders or probationers" (#19247). Tier III is a five-wing, five-phase program with a stated 13-month (390-day) minimum — at least 60 days in E-Wing, 60 in F-Wing, and 90 each in D-, C-, and B-Wings — and a stated 24-month maximum (#19260). Tier placement is formally tied to security classification: all Tier III, Tier II STEP, and Tier III STEP prisoners are classified Close Security (#19262).

The state has quantified this system exactly once and then stopped. GDC's FY2014 Annual Fiscal Report states that, as part of the Tier Segregation Management System installed at all Level V facilities, "2,721 beds were converted to Tier beds" — a bed count, not a population count, and the only system-wide quantification GDC has ever published (#19297). Four years later it reported reductions rather than levels: FY2018 reported a 41 percent decrease (671 people) in the extended restrictive housing population since January 2017, and FY2019 reported a 40 percent decrease (675) from the same baseline — two statements that cannot be reconciled to a single baseline, and which supply a delta with no endpoints (#19298, #19299). Restrictive-housing and Tier programme narrative appears in the annual reports for FY2014, FY2017, FY2018, FY2019 and FY2020 and then disappears: from FY2021 to FY2025 there is no published figure, percentage or trend for the restrictive-housing population, and the last year in which GDC published any number about its size or direction is FY2019 (#19303). A systematic review of 110 non-monthly GDC reports and all 34 distributions of the monthly Inmate Statistical Profile series found no published Tier or restrictive-housing population count anywhere — and no GDC policy that requires one to be compiled (#19306, #19308). The reason is not that the data does not exist: GDC's own annual reports record that a Tier II SCRIBE module went live in March 2019 "to effectively track offenders in Extended Restrictive Housing," and that the FY2020 SCRIBE Warden's Dashboard displays, for Tier II offenders, "number of days an offender has spent in restrictive housing Tier II" (#19300). The data exists in an electronic system of record; no policy instrument requires anyone to aggregate it, because every Tier approval travels as a single case — Classification Committee to Warden for Tier II, and facility Warden to Tier III Committee to SMU Warden to North Regional Director to Director of Field Operations for Tier III (#19308).

The one Tier population GDC does publish is a side effect of the Special Management Unit being a standalone facility with its own GDC facility code. The SMU appears as a facility row in the monthly Inmate Statistical Profile's population table; as of 2026-09-01 it held 154 people, and across 75 monthly snapshots from 2020-07-01 to 2026-09-01 its population ranged from 149 to 195 (#19304, #19376). Against its stated 192 single-bunked capacity, that implies the unit has run below bed capacity throughout the published series (#19302, #19484). It is Tier III at one facility. It says nothing about Tier I or Tier II, whose units sit inside ordinary prisons and are absorbed invisibly into those prisons' facility totals (#19304). GDC's only per-facility Tier cell figure is likewise a single facility: FY2018 recorded that Guard 1 Technology at Valdosta State Prison "captures automated segregation rounds in all tier units (178 Cells)" (#19301). One prison's Tier units alone approach the size of the entire SMU.

Who Decides, and Where the Authority Sits

Tier II placement is decided by facility staff; release is not. SOP 209.08 (effective 04/11/2016) constitutes the Tier II Classification Committee of the Unit Manager, the Tier II OIC, the assigned counselor and, as applicable, the assigned mental-health counselor; the Warden approves or disapproves all recommendations within seven business days, and in an emergency the Warden may authorise immediate assignment (#19256). The exit chain runs the other way: Unit Manager recommendation, Warden and Regional Director review, a Criminal Investigations Unit Security Threat Group clearance within ten business days, then an electronic transfer to the Director of Field Operations, who "will make the final decision," after which a statewide Tier Segregation Manager reviews the record in SCRIBE and "determine[s] the appropriate placement of the offender" (#19256, #19257). A facility can put a person into Tier II on its own authority, including immediately; it cannot take a person out. That asymmetry is written into the policy.

The Tier Segregation Manager is a single statewide position whose function is to place people within the Tier system, and is the strongest documentary indication in the SOP corpus that a centralised Tier placement record exists (#19257). Training runs the other way: SOP 506.08 creates a centrally delivered SMU/Tier III training programme for officers assigned to those units and no corresponding Tier I or Tier II programme, leaving training for the multi-site Tier I and Tier II programmes to each Warden, with SOP 209.07 promising in a policy effective in 2015 that "Specialized departmental training will be developed" (#19290).

The review chain is also one-directional where it counts. The appeal of initial Tier II assignment goes to the Director of Field Operations; the appeal of the 90-day review — the decision that continues confinement — goes only to the Warden, who is also the official who approved the placement in the first instance (#19258). The first 96 hours of administrative segregation rest on a placement decision that SOP 209.06 makes expressly unappealable, and the appeal that does exist terminates with the Warden, whose "review and decision on the appeal is final" (#19294).

Admission Without Adjudication

Tier I lists eight grounds for placement, only one of which presupposes a pending disciplinary charge, and none of which requires a disciplinary conviction. Three — a staff determination that the person is "noted as a threat to the safe and secure operation of the Facility," that the person is "the subject of an investigation regarding a serious violation," or that the person "has not been classified due to their arrival at a Facility" — permit segregation on staff determination alone (#19248). Tier II's broadest admission criterion — "noted as a threat to the safe and secure operation of the Facility" — is open-ended on its face, carries no evidentiary standard, and expressly covers a person who "ha[s] been threatened with bodily harm" (#19253). Criteria 1, 5 and 10 (threat, failure in Tier I, assaultive histories) can each be satisfied without any disciplinary adjudication (#19253). Tier II's criteria also establish two further facts: that a "Tier II-Mental Health" programme exists as a distinct placement, and that the Special Management Unit discharges people into Tier II Phase 3 rather than to general population (#19255). The GDC Fact Sheet quoted in five PREA audits adds two more grounds worth naming: refusal to participate in Tier I is itself a ground for escalation to Tier II ("Failure in Tier I or refused to participate"), and transfer from Georgia Diagnostic and Classification Prison is a standalone eligibility ground for long-term administrative segregation (#19391, #19487, #19488).

GDC's framing does not match that drafting. SOP 209.08 states that Tier II "is an offender management process and is not a punishment measure," and SOP 209.06's summary states that segregation "is not a form of abuse or punishment" (#19291). But SOP 209.06's own placement grounds include "Serving sanctions for a violation of Facility/Center rules," and SOP 209.01 Section W caps disciplinary isolation at 30 days in prisons and transitional centres and 14 days in probation detention centres and integrated treatment facilities while designating a Tier II recommendation as the route when isolation time exceeds 30 days — that is, Tier II is written into the discipline SOP as the mechanism for continuing confinement after the sanction cap is reached, in a section whose stated premise is a finding of guilt (#19292, #19293, #19352). A quoted source in GPS's 2026 material describes the unit in the same terms the policy disclaims: "The tier is a special lockdown unit in several Georgia prisons. It's for major punishment," adding that "too often the prison administrator just uses it to get someone out of the way — hide them" (#19317, #19318). That quotation is reported, not a finding.

The Conditions Floor, and What It Adds Up To

Georgia's Tier policies guarantee startlingly little out-of-cell time, and the arithmetic is GDC's own. Tier I's floor is five hours per week across five days, at least one hour per day on five days, leaving two days a week on which the policy guarantees no out-of-cell time at all (#19270). Tier II's floor is also five hours per week, with no daily distribution specified — the five hours may lawfully be delivered in any pattern across the week (#19269). Generic administrative segregation under SOP 209.06 (2021) provides one hour per day, five days a week (#19271). The juvenile programme's floor is identical to adult Tier II: five hours per week, yielding the same 23.3 hours per day in cell for a person under 18, with the out-of-cell mental-health group entitlement expressly conditioned on "availability of security staff" (#19266). Tier II STEP and Tier III STEP guarantee four hours out of cell per day, seven days a week (#19283, #19284); Tier III guarantees four hours out of cell per day, Monday through Friday (#19281).

Derived by arithmetic from those floors — a figure that states what GDC policy guarantees as a floor, not what any individual experiences or an observed average (#19272): Tier I, Tier II, generic administrative segregation and the juvenile programme each work out to 163 hours per week in cell, or 23.3 hours per day on average, with 24 hours in cell on the two days a week carrying no guaranteed out-of-cell time. Tier III averages 21.1 hours per day on weekdays but 24 hours on Saturdays and Sundays, when no out-of-cell time is guaranteed. Tier II STEP and Tier III STEP, at 20.0 hours per day, sit below the 22-hour line. The effect is that the stage GDC labels most restrictive guarantees four times more out-of-cell time than Tier II, and the step-down programmes guarantee 5.6 times more (#19273). Set against the UN Standard Minimum Rules for the Treatment of Prisoners (the Nelson Mandela Rules, adopted 17 December 2015), whose Rule 44 defines solitary confinement as confinement for 22 hours or more per day without meaningful human contact and prolonged solitary confinement as confinement in excess of 15 consecutive days, and whose Rule 43 prohibits both indefinite and prolonged solitary confinement in all circumstances: on GDC's own written minimums, Tier I, Tier II, generic administrative segregation and the programme for children all meet the 22-hour threshold as a policy floor every day of the week, while the reformed Tier III sits below it on weekdays (#19274, #19329). The Mandela Rules are persuasive and are not enforceable in United States courts (#19329).

Two caveats keep this honest. First, a policy floor is not an observed average (#19272). Second, whether the confinement is "solitary" in the Rule 44 sense turns on the element GDC does not specify: absence of meaningful human contact. SOP 209.09 and SOP 209.55 guarantee single occupancy in Tier III and Tier III STEP; SOP 209.45 permits double occupancy in Tier II STEP; SOP 209.08 is silent on occupancy in Tier II, and SOP 209.06 Section IV.G treats single-celling in administrative segregation as an enumerated exception requiring assignment by medical, mental-health or classification staff plus evaluation within 24 or 48 hours — implying that multiple occupancy is the ordinary condition (#19278, #19279, #19331). Where a Tier II cell is single-occupancy both elements may be satisfied; where it is double-bunked the person is in cell 23.3 hours a day but is not without human contact. Whether a person in Georgia's Tier system is held alone or with a cellmate is determined by which unit they are assigned to, and GDC's Tier II policy neither specifies nor reports it (#19510). The audit record confirms double-celled Tier housing as ordinary: Hancock's G-1 "Tier 2" unit houses "a maximum of 96 inmates, in double occupancy cells," in identical language across two audit cycles, and Hancock's J Building Tier 1/segregation unit holds 20 double-occupancy cells (#19572, #19573). Across nine facilities, GDC's own audits describe double-bunked segregation and Tier housing as the ordinary described condition — Valdosta's Tier 1 in "double occupancy cells," Telfair's E-1 and E-2 holding 80 and 70 "in two man cells," Smith's dorms of "48 double bunked cells housing 96 offenders per dorm," Ware's J-Building holding 100 Tier 1 offenders double-bunked, Macon's "48 double occupancy cells of segregation beds," and Georgia State Prison's Tier II at G Building double-bunked while Tier II at K Building was single-cell lockdown (#19395, #19528). At Hays, Tier I units were two-man cells and Tier II units one-man cells (#19522). The two occupancy modes run inside the same programme at the same prison (#19529).

The conditions themselves, in GDC's own words, are thin. SOP 209.08 requires cells "equipped and furnished in a manner consistent with cells designated as Tier II cells" — a circular standard, where Tier III, Tier II STEP and Tier III STEP all instead require furnishing "consistent with cells in the general population" (#19275). Personal hygiene is offered three times per week; food is "the same quality and quantity as that provided in the general population"; exercise is "a minimum of five (5) hours per week"; legal access is as provided in SOP 227.03; visits are non-contact only, one to three per month by phase (#19275). Tier II's religious-practice clause lets offenders pray "in their assigned cells or by their assigned beds" — phrasing that only has work to do if some Tier II housing is not single-cell (#19277). And a single clause lets the Warden modify every one of those conditions: "As deemed necessary by the Warden/Designee, all of the above conditions or privileges are subject to modification due to extenuating operational circumstances" — a term the SOP does not define, with no time limit, no documentation requirement, no notice to the prisoner, and no reporting obligation to Central Office (#19276). Tier III (2025) requires comparable exceptions to be "recorded in the Electronic Logbook ... and justified in writing," and requires written notice to the North Region Director whenever out-of-cell time is cancelled or shortened for three or more consecutive days; Tier II has no equivalent (#19276, #19282). The tablet is confiscated on Tier II placement until Phase III, and video visitation in Phase III is capped at one visit per calendar month within a six-hour window on two named weekdays; the STEP programmes and Tier III instead guarantee general-population device access (#19285).

Two further conflicts sit inside GDC's own medical and clinical rules. SOP 507.04.33 (2022) requires daily medical rounds in restrictive housing; Tier I SOP 209.07 (2015) and Tier II SOP 209.08 (2016) require medical rounds three times per week, expressly excluding weekends and holidays — and the older, less protective standard is the one written into the policy that governs the Tier unit itself (#19334, #19335, #19336, #19337). Board Rule 125-4-4-.08 likewise requires medical checks three times weekly, creating a conflict with SOP 507.04.33's daily standard (#19354). Only the Institutional Duty Officer is required to attend on weekends and holidays; health care, the Chief of Security, the Unit Manager and the Deputy Warden are all expressly excused (#19340, #19341, #19342). On mental health, GDC's own clinical policy ranks Tier II among its highest-need units: SOP 508.01 (2023) places "SMHTUs/TIER II" in the top priority tier of its staffing hierarchy alongside death row, the SMU and acute care, and assigns them a 1:20 counselor ratio — behind only crisis stabilization at 1:15 (#19288). The same policy names a SCRIBE "Daily lock down" report as a routine mental-health oversight instrument, which undercuts any claim that per-unit confinement data cannot be produced (#19289). Suicide-risk questioning is monthly in Tier I and weekly in Tier II — the unit with the shorter maximum stay carries the less frequent screening — and the SOP says offenders "should be asked," not "shall" (#19287). Tier II admits prisoners at mental-health Level III on mental-health recommendation and excludes only Level IV; Tier III excludes both Level III and Level IV, routing the higher-need prisoners to the unit with the five-hours-per-week floor rather than the one reformed under litigation (#19254).

Duration

Duration is bounded inconsistently. Tier I is capped at 30 days except for protective custody, voluntary or involuntary, which is uncapped (#19249) — the only outer time limit anywhere in the adult Tier architecture, and one disapplied for a category staff can impose involuntarily. Tier II has no outer limit written into policy: SOP 209.08 provides that offenders who "cannot return to general population" remain in Phase 3 subject only to a 90-day review that may confirm retention, and the SOP contains no maximum duration (#19252). Tier III is a stated 13-month (390-day) minimum and a stated 24-month maximum, with six enumerated criteria permitting retention past 24 months (#19260). The 24-month ceiling is stated and then immediately qualified — "the Department reserves the right to retain offenders in the Tier III Program for a longer period of time" — so it operates as a presumption subject to Commissioner-level override rather than as a cap, though it is a presumption with a named decision-maker (a four-member statewide panel of the Director of Field Operations, the Statewide Mental Health Director, the Statewide Medical Director and a member of the Office of Legal Services), a named form (Attachment 12), a fixed quarterly review interval, and the requirement of personal approval by the Commissioner or Assistant Commissioner for Facilities (#19261, #19330).

Setting those durations against the Mandela Rules' 15-day prolongation threshold: Tier I's 30-day cap is 2.0 times the threshold, and unbounded for protective custody. Tier II's 90-day review cycle is itself 6 times the threshold and indefinitely repeatable, with no maximum duration stated. Tier III's 390-day minimum is 26.0 times the threshold as a floor — a person who progresses successfully through every phase still serves 390 days — and its roughly 730-day outer limit is approximately 48.7 times it (#19329).

What Was Won, and Where It Stopped

The central scope fact about the Gumm litigation is that it reaches Tier III and the SMU only, not Tier II. Every operative settlement term is written in SMU and Tier III vocabulary: a general 24-month limit on SMU confinement, a prohibition on housing Mental Health Level III or above in Tier III/SMU, mandatory 60-day and 90-day mental-health evaluations, the quarterly four-member panel, and Commissioner-level approval to hold anyone past 24 months. GDC codified all of it into SOP 209.09 alone, revised effective 2025-04-23 (#19324). SOP 209.08, governing Tier II, has none of it and has not been revised since 2016-04-11 — nearly three years before the settlement. No duration cap. No mental-health exclusion. No panel. No mental-health evaluation requirement at placement. No out-of-cell minimum beyond five hours per week (#19324). The programme with the least protection is the larger one.

The Gumm record itself supplies the sharpest conditions evidence in the corpus, and it is about the SMU. The court adopted findings, based on the Dr. Craig Haney expert report, that SMU residents were confined in cells "smaller than the average parking space," with as little as five hours per week out of cell, no outside light, no congregate religious or educational programming, and frequently locked food-port flaps; at the 2017 inspection the SMU held approximately 180 people, 70 of them designated mentally ill. Timothy Gumm was held continuously for seven and a half years despite fourteen transfer recommendations over four years, and people were released directly from the SMU to the community at sentence expiration with no transitional programming (#19326). Haney's quoted opinion was that the SMU was "one of the harshest and most draconian" he had seen "in decades of conducting evaluations," and that the harm "may be irreversible and even fatal" (#19326). A prior GPS datapoint described the Haney findings as applying to "Tier II/SMU residents"; the Haney inspection and the certified class were the Special Management Unit at Georgia Diagnostic and Classification Prison, not Tier II, and the Gumm record supplies no findings about Tier II conditions at any other facility (#19325). The distinction is material: the five-hours-per-week figure in the Gumm record is a 2017 SMU finding, and it happens to coincide with the figure SOP 209.08 still sets as the Tier II policy floor today — but they are two separate facts about two different programmes (#19325).

A separate 2024 order shows what enforcement of even the narrow settlement produced. On April 19, 2024, Chief Judge Treadwell found that GDC had "no desire or intention" to comply with the December 2018 settlement, and ordered an independent monitor, fines of $2,500 per day (amounting to $75,000 every 30 days for six months), and additional attorney's fees, extending the settlement past its initial three-year term. The order also documents that officials falsified therapy records, including recording that a man attended treatment sessions after he was already dead (#19322).

Wilkinson v. Austin, 545 U.S. 209 (2005), points the same way as the scope finding: the factor the Court identified as central to a due-process liberty interest — indefinite duration — is present in Tier II, where SOP 209.08 sets no outer limit, and absent from Tier III, where SOP 209.09 sets a 24-month ceiling with a quarterly panel. On Wilkinson's own reasoning the due-process argument is stronger for Tier II than for the programme the litigation actually reformed. That is an argument from a holding, not a holding; no court has applied Wilkinson to Georgia's Tier II programme, and Sandin v. Conner, the decision Wilkinson refines, is absent from the GPS research library (#19327). GPS also holds no document establishing that Georgia's Tier programmes are used to retaliate against protected activity, and no court has made such a finding; the case-law corpus's silence is not itself evidence of absence (#19312).

How Many, and Why No One Can Say

The published data structure makes the Tier population arithmetically unrecoverable rather than merely unpublished. SOP 209.09 classifies all Tier III offenders as Close Security, and SOP 209.45 and SOP 209.55 do the same for the STEP populations; the monthly Inmate Statistical Profile's only housing-adjacent field is "supervision," which reports Close, Medium and Minimum. The Tier population is contained within GDC's published Close Security count — 11,600-plus people — but cannot be isolated from it, because GDC publishes security classification and not housing status (#19307).

The best per-facility segregation dataset GDC publishes is its PREA audit corpus, and it is internally inconsistent. A scan of all 273 GDC-published PREA facility audit reports (Cycles 1 to 5) found documented Tier programme housing — Tier I, Tier II, Tier III, "Tier Program dorms," "Tier units" — in the audits of exactly nine facilities: Georgia State Prison, Smith, Hays, Valdosta, Telfair, Macon, Hancock, Georgia Diagnostic and Classification State Prison, and Ware. All nine are close-security prisons, with not one exception in 273 documents (#19359). The nine are the seven facilities GDC designated as Tier I and II — Hancock, Hays, Macon, Smith, Telfair, Valdosta and Ware — plus Georgia State Prison (closed 2022-02-19) and GDCP, which hosts Tier III, the Special Management Unit (#19359, #19369). Several audits state the linkage in terms: Valdosta Cycle 2 — "In addition to it being a close security prison, it also houses specialized units called Tier Programs"; Smith Cycle 2 — "The facility is also a Tier 1 and Tier II facility ... primarily housing close security offenders"; Telfair Cycle 2 — "The facility is a Tier I and Tier II Facility" (#19360). Of the twelve currently active close-security state prisons, eight have documented Tier programme housing; the three whose audits contain no Tier language are Baldwin State Prison (K1, K2 and K3 described as "Isolation/Segregation," never Tier), Phillips State Prison (100 isolation/segregation cells with no Tier language), and Augusta State Medical Prison (#19361, #19372, #19373). All seven designated Tier I and II facilities are men's prisons; GDC's two close-security-classified women's facilities, McRae Women's Facility and Arrendale State Prison (Mixed Close/Medium), are not on the designation list, and whether GDC operates a Tier II programme for women at all is not answered anywhere in the SOP corpus or the published fact sheets GPS holds (#19364, #19365). The seven-facility designation itself derives from GDC fact sheets of which GPS holds only the 2018 and 2019 editions, and should be cited as GDC's designation as of 2018-2019, not as the current set: Georgia State Prison closed in 2022, SOP 209.09 was revised effective 2025-04-23, and SOP 508.01 (2023) refers to "designated SMHTU/TIER II facilities" without naming them (#19366, #19370). A keyword search of the 1,432-document SOP corpus for a Tier facility designation list returns nothing; the current designation document is the single most load-bearing missing document for the question of Tier's system-wide scale (#19264).

But segregation cells are not confined to the Tier designation. GDC's PREA audits report segregation counts at medium-security prisons throughout the system, sometimes larger than at close-security prisons: Calhoun State Prison, whose Cycle 2 audit gives its security level as "4/Medium," reports 197 segregation cells — more than close-security Baldwin (167), Ware (125) or Phillips (75) — and Wilcox State Prison, described in its own audits as "housing medium security, general population inmates," reports 168 in Cycle 3, up from 24 double-bunked in Cycle 1 and 72 in Cycle 2 (#19399, #19442, #19467). Coffee Correctional Facility, a private medium-security prison, reports 170 (#19444). Every state prison in the corpus that reports a number reports segregation cells, close security or medium (#19511). The Tier label marks where GDC concentrates long-term administrative segregation programming; it does not mark where segregation cells exist (#19512).

Summing the most recent audit figure per facility that states one gives at least 3,401 segregation cells across 30 facilities — a floor, not a total, and a cell count, not a bed count or a population (#19398). Restricting to state-operated facilities gives 3,182 across 28; on a single-cycle basis, the 23 state facilities with a structured Cycle 2 Pre-Audit Questionnaire figure total 2,967, rising to 3,078 across 25 when Wilcox and Montgomery are added from narrative (#19504, #19505, #19506). Four caveats travel with the figure. It is a cell count: the audits are explicit that many of these cells hold two people — Lee reports "16 cells with 32 beds," Wilcox "72 double bunked segregation cells," Montgomery "eight (8) two-man segregation cells," Macon "48 double occupancy cells of segregation beds" — so the bed count is materially higher and the number of people is a separate unknown GPS has not interpolated (#19398, #19507). It is a floor, not a total: at least six state facilities with published audits contribute zero because no audit of them ever states a figure — Arrendale State Prison (three audits), Augusta State Medical Prison (four audits), Walker State Prison (five audits), Helms Facility, Emanuel Unit, and McRae Women's Facility (no audit in the corpus under that name) (#19415). It mixes vintages spanning Cycle 2 (2017-2019), Cycle 3 (2019-2021) and Cycle 4 (2021-2025), and the most recent audits of five of the nine Tier prisons — GDCP, Hays, Smith, Macon and Wilcox — dropped the segregation-cell field, so those facilities contribute older numbers (#19416, #19508). And it counts all segregation, not only Tier programme housing; Tier units are a subset (#19507). It must not be presented as a trend against the FY2014 figure of 2,721 Tier beds converted at Level V facilities: the two are different measures a decade apart (#19398, #19417).

The corpus's data-integrity defects are themselves part of the finding, and any published figure drawn from the audits must name the facility, the cycle and the field, and must not silently reconcile a questionnaire figure against a narrative one (#19400, #19513). Georgia State Prison's "Cycle 3" audit is the Cycle 2 audit republished — both 181 pages, identical extracted text, zero differences, both dated 2019-03-22 — so its segregation figure must not be counted twice (#19400, #19408). Emanuel Women's Facility's Cycle 5 audit is published twice identically (#19409). Telfair State Prison's Cycle 4 audit contradicts itself internally: "The PCM indicated the facility does not have administrative segregation cells," while the same report's interviews section records "360 administrative segregation beds" (#19400). Pre-Audit Questionnaire segregation fields contradict the same audit's own narrative in at least three cases — Georgia State Prison Cycle 2's questionnaire says 12 segregation cells while its narrative itemises 106 Tier I cells at E-South, 218 Tier II beds at G Building and 100 Tier II single lockdown cells at K Building (#19411). The same "does not have administrative segregation cells" form response appears in Cycle 4 and 5 audits of prisons whose earlier audits reported substantial segregation capacity (#19400). Label defects leave the unit of measure undetermined in the Washington, Dooly and Pulaski audits (#19413). The word "tier" in a PREA audit usually means the upper or lower gallery of a cellblock, not the Tier programme (#19401), and a single sentence of GDC boilerplate about delayed PREA education mentioning "Tier Program" appears in roughly 60 audits, including audits of facilities that demonstrably have no Tier housing, so a naive keyword search produces a badly inflated facility list (#19402). One facility deserves its own caution: Augusta State Medical Prison's "50 Tier II Mental Health beds" is ambiguous — in GDC usage "Level II" is also a mental-health classification — and Augusta is not counted as a Tier-housing facility (#19400). GDC's PREA audit URLs are also not uniformly patterned, and should never be constructed by hand; the source_url and citation string must be taken from the report index for the specific facility and cycle (#19571).

Reading the audits for what they do say about scale, the unit-level figures are large. Hays State Prison Cycle 2 (report dated 2019-06-15) states 303 Tier beds across its Tier I and II units — D2 and E at 64 each in two-man cells, F1, F2, G1 and G2 at 32 one-man cells each, and H1 and H2, named as "Tier 2 Stepdown Unit," at 47 — larger than the entire Tier III Special Management Unit (#19393, #19472, #19523). Georgia State Prison Cycle 1 (2016) comprised "17 general population dorms and 13 Tier Program dorms," and its Cycle 2 states 424 Tier beds — E-South 106 double-bunked, G Building 218 double-bunked, K Building 100 single-cell lockdown — more than twice the SMU's 192 (#19394, #19473, #19474, #19475, #19524). The Southern Center for Human Rights' September 2021 complaint alleged approximately 300 people in Georgia State Prison's Tier II programme, over 70 percent of them with serious mental illness; that is an allegation in a complaint, not a finding, and describes a facility that has since closed, but it is the only facility-level Tier II population figure in any source GPS holds (#19367, #19328). Set one prison's roughly 300 against the entire SMU's published 149 to 195 and the comparison — flagged as derived, and not to be extrapolated to a system total — shows why a system-wide Tier II count matters more than the SMU figure GDC does publish (#19368).

Duration figures for Tier II appear only in audits, never in the SOP: Hays describes Tier II as "long-term segregation/270 days," Macon's Cycle 2 and Cycle 4 audits describe it as running "from nine months to indefinite" and "from 270 days to an indefinite date," and Georgia State Prison's Cycle 2 records "up to two years" for Tier II progression. SOP 209.08 states no duration at all (#19403, #19494, #19496, #19497, #19478). The audits also record the double-bunking arithmetic directly: Smith's dorms comprise "48 double bunked cells housing 96 offenders per dorm," and Hays shows the same pattern with Tier II one-man dorms at 32 and Tier I two-man dorms at 64 (#19526, #19527). Unit capacities run from 20 to 218 per unit; there is no standard Tier dorm size (#19509).

The Records That Exist, and What to Ask For

The absence of a published count is not an absence of records. SOP 209.07 and SOP 209.08 each require per-cell, per-bed "Accountability Logs" recording "cell number, bed number, offender name, offender number, race, in or out time, assignment of offender and destination of offender" — by their own description a daily record of who is housed in each Tier cell and bed, and the facility-level record that would establish Tier population and occupancy mode (#19296). SOP 209.06 requires a segregation admissions and releases register recording name and GDC number, housing location, date and time of action, reason for admission and tentative release date, and authorising official or committee — yielding both population and length of stay (#19309). Board Rule 125-3-2-.10(f) requires Division Director review and approval for any isolation beyond 30 days, so every long-stay confinement generates a dated Central Office approval record; counting those for a defined period gives a floor on long-stay isolation independent of any Tier population figure (#19350, #19351). SOP 209.09 Attachment 12 — the quarterly over-24-month review form — plus the Commissioner's personal approvals, yields the number held in Tier III beyond 24 months (#19309). The Restrictive Housing Rounds Log (form M40-01-01), the Sick Call Encounter Log for Restrictive Housing (form P-03-0005-11), and the SMU Warden's written notices to the North Region Director whenever out-of-cell time is cancelled for three or more consecutive days are all named, existing records (#19309, #19282). The 30-minute watch forms and the colour-coded sign-in logs — Wardens and Supervisors in red ink, Health Care in green, Counselors in blue, Officers in black — are self-auditing as to which category of official attended and when, and GDC's own prohibition on completing them in a batch ("gang" entries "shall never be permitted") is evidence that the practice was known (#19344, #19345, #19346, #19347, #19348, #19349). The 30-minute/15-minute checks are documented "when they occur and not at the end of the shift" (#19347). The forms and logs are enumerated in the SOPs down to their attachment numbers: six Tier I forms, eleven Tier II forms, and the Tier III quarterly review form (#19296). And the record that would answer the Tier retaliation question is specific: the SCRIBE Tier assignment record and the Classification Committee 90-day review forms set against that person's grievance filing dates — the sequencing of those two documents is the entire question — or, in aggregate, Tier II and Tier III placements by facility and month cross-tabulated against grievance filings, and the Director of Field Operations' Tier release decisions with stated reasons (#19314).

GDC's standard obstacle is not that it lacks the data but that it treats the records as confidential. SOP 219.01 asserts that "Offender records are confidential state secrets under O.C.G.A. § 42-5-36(c)" and directs all requests for offender-related records immediately to the GDC Office of Legal Services; a request framed around individual records will meet that exemption, while a request framed around aggregate counts, unit rosters stripped of names, and designation and capacity documents will not (#19296). GDC's standard objection — that compiling a count creates a new record — is answered by the Accountability Log being printed daily by policy and by the agency's own annual reports establishing that the SCRIBE Tier II module exists and the Warden's Dashboard displays days served (#19300, #19309). The request must also ask across the whole vocabulary — Tier I, Tier II, Tier III, extended restrictive housing, restrictive housing, administrative segregation, disciplinary isolation and protective custody — because GDC's terminology drifts by year and because SOP 209.45 and SOP 209.55 state affirmatively that "The Tier II STEP is not a restrictive housing unit" and "The Tier III STEP is not a restrictive housing unit," a definitional carve-out that will shrink any production framed on a single term (#19309, #19333). The precedent is favourable: in open-records request R024615-101925, filed 2025-10-19, GDC produced per-facility count tables of inmates by security designation at each state prison, demonstrating it will run and produce a per-facility housing count query, which undercuts a burden or new-record objection; no GPS request to date has asked for Tier or segregation population data (#19310).

Two documented conflicts should be recorded rather than resolved, because each bears on what the records will show. SOP 209.06 (Administrative Segregation, 2021) states that offenders "shall have the same access to GOAL Devices and the J-Pay Kiosk as that of the general population" and that general-population visitation privileges "shall be allowed," while SOP 204.10 (2018) confiscates the GOAL device on Tier II placement until Phase III and SOP 209.08 (2016) limits Tier II visits to one to three non-contact visits per month; both the generic policy and the Tier-specific policy claim the field, and the corpus does not resolve which governs (#19286). And SOP 209.05 (Stripped Cells, 2015) limits stripped-cell placement to emergencies, never punishment, with written Warden authorisation, an eight-hour initial maximum and daily medical authorisation to continue, while SOP 209.08 imports that policy only for the narrow question of bedding supplies, sets no time limit of its own on moderated-cell placement, and lists as a ground for it "Continuous violations of policy, rules, & regulations while assigned to Tier II" — which is not an emergency and is punishment-shaped (#19383, #19384). GDC policy also recognises that the behaviours leading to a stripped cell are frequently psychiatric in origin and requires a clinical determination before or alongside the security response; whether that determination is made and recorded in practice is unknown, and the documentation required by SOP 209.05 for each stripped-cell placement is the record that would show it (#19386, #19387, #19388, #19389, #19390).