Deaths in Custody

Brief written August 9, 2026 from GPS Intelligence System data.

Georgia's state prison system, which holds approximately 53,000 people across 38 facilities, has become one of the deadliest in the United States. In October 2024, the U.S. Department of Justice published a 93-page findings report determining that the Georgia Department of Corrections violates the Eighth Amendment's prohibition on cruel and unusual punishment. The report, which the DOJ described as revealing "among the most severe violations" uncovered in any prison investigation, found that GDC leadership has "lost control of its facilities," that incarcerated people endure "near-constant life-threatening violence," and that the State is "deliberately indifferent" to the risk. But the public death numbers alone — the observable surface of the crisis — tell only a fraction of the story. The system's own machinery for counting, classifying, and disclosing how people die has broken down so thoroughly that for most of the deaths in Georgia's prisons, no one outside the state can say with certainty what killed them.

The Scale of Death

The raw figures are staggering. Georgia Prisoners' Speak has independently tracked 1,860 deaths in GDC custody since 2020. GPS's mortality database, which cross-validates its annual totals against UCLA Law's prison mortality project, logged 294 deaths in 2020, 257 in 2021, 254 in 2022, 262 in 2023, and 333 in 2024 — GPS's 2024 count exactly matching UCLA's figure. Among the deaths GPS has been able to classify, 644 were attributed to natural or medical causes, but 258 were homicides, 110 were suicides, 49 were overdoses, and 8 were categorized by GPS as deaths from medical neglect. Critically, the single largest category — 720 deaths, or nearly 39% of the total — carries a cause of death listed as "Unknown/Pending." That number is itself a product of the system this article examines.

The homicide count alone marks a profound departure from historical norms. Between 2001 and 2019, the Bureau of Justice Statistics' now-defunct Mortality in Correctional Institutions program recorded only 89 Georgia prisoner homicides cumulatively. By contrast, the DOJ documented 142 homicides in the six-year window from 2018 through 2023 — a 95.8% increase from the first three years to the last three — with the annual count climbing from 7 in 2018 to 35 in 2023. In the first five months of 2024, GDC incident reports showed at least 18 confirmed or suspected homicides, though GDC publicly reported only 6 in its mortality data for that period.

Suicides, too, have quadrupled. GPS's analysis of federal data, reported in the publication Suicide in Georgia State Prisons: The Documented Rise (2001–2021) and the Reporting Blackout After It, found that Georgia's prison suicide rate rose from 8 per 100,000 prisoners in 2010–2014 to 31 per 100,000 in 2015–2019 — a rate that exceeded the U.S. total, the all-states figure, and the Southern regional average. That rise was specific to GDC; Georgia's county jail suicide rate remained flat over the same period, ruling out a general statewide trend. Then, after 2019, the federal data series ended. The Death in Custody Reporting Act transferred collection to the Bureau of Justice Assistance, which has not published the data. From 2022 onward, Georgia's own death records carry no cause of death at all.

How the State Conceals How People Die

The opacity is not accidental. In March 2024, GDC announced it would no longer provide cause-of-death information in its monthly mortality reports, stating it would release manner of death only after local medical examiners make their determinations. That process can take a year or more — and for many deaths, a final determination never arrives. GPS's investigation Who Decides How They Died: Georgia's Elected-Coroner System, Medical Examiner Capacity, and the Structural Roots of Undisclosed Causes of Death traced the structural reasons. In 155 of Georgia's 159 counties, the official who takes charge of a body and certifies a death is an elected coroner who needs no medical training — only a high school diploma, a short course, and registered voter status. Only four counties (Fulton, Cobb, DeKalb, Gwinnett) have replaced the coroner with a physician-led medical examiner.

When a person dies inside a state prison, the coroner of the county where the fatal act occurred — not necessarily the county where the hospital is — takes jurisdiction. That coroner may order an autopsy, but the decision is discretionary and limited by GBI capacity. The GBI Medical Examiner's Office, which serves most counties, averages roughly 300 autopsies per pathologist per year, exceeding the National Association of Medical Examiners' recommended maximum of 250 and approaching the hard accreditation cap of 325. As GBI leadership has stated, the pace is not sustainable. A coroner may certify a death without an autopsy and may enter the cause as "pending" indefinitely. In many rural counties, the coroner is part-time, per-case compensated, and without the forensic pathology expertise to flag that a death classified as "natural" may, in fact, have been driven by chronic undernutrition, neglect, or violence. The 2010 state audit had already found that the State Medical Examiner's Office was not contacted in 26% of statutorily required suicide referrals.

The consequence is a chain of documented weaknesses: lay coroner certification, GDC's control of notification and scene, discretionary autopsy limited by capacity, funds, and coroner judgment, and a cause of death that sits pending while records remain closed. GPS's investigation detailed how GDC internal investigation reports are "classified as confidential state secrets" under state regulation, and how the Department's own death investigations under SOP 507.04.67 are separate from, and do not necessarily reconcile with, the certified cause. The result: for 720 of the deaths GPS has tracked, no one outside the state can verify the cause.

The DOJ Findings: Deliberate Indifference as a Matter of Record

The Department of Justice's October 2024 findings report was the culmination of a nearly three-year investigation that produced more than 19,000 records — a process DOJ described as "unnecessarily contentious" and which required a subpoena to compel document production. The findings presented a systemic indictment: 82 recommended remedial measures; a documented causal chain linking chronic understaffing, classification breakdown, gang control of housing units, and near-constant violence; and a finding that GDC had obstructed investigators and conducted pre-inspection cleanups. Assistant Attorney General Kristen Clarke summarized: "In America, time in prison should not be a sentence to death, torture or rape."

Among the specific findings: correctional officer vacancy rates exceeding 70% at eight facilities, systemwide vacancies over 50% since mid-2021, with single officers sometimes responsible for supervising 400 or more beds. The DOJ found that classification decisions were driven by bed availability rather than risk assessment, that close-security inmates were housed in medium-security facilities not staffed for that population, and that GDC does not adequately screen, classify, or track LGBTI individuals, who face heightened risk of sexual violence. Violence was consistently underreported and mischaracterized; incidents were coded in ways that obscured their true nature. GDC publicly rejected the findings as reflecting a "fundamental misunderstanding," but in a separate federal litigation, Judge Marc Treadwell had already held GDC in contempt for falsified reporting, observing that "the Court has long passed the point where it can assume that even sworn statements from the defendants are truthful." That contempt order cited a discrepancy between GPS's count of 100 deaths for part of 2024 and GDC's official count of 66.

The Drivers: Staffing Collapse, Classification Erosion, and Undernutrition

The DOJ's causal chain begins with staffing. Georgia's state prison census has doubled since 1990, while correctional officer staffing stands at roughly 50% of full levels systemwide — and below 30% at some facilities. The vacancy rate averaged 49.3% in 2021, 56.3% in 2022, and 52.5% in 2023, peaking at 60% in April 2023. Twelve prisons have had vacancy rates above 70%. This understaffing breaks the basic mechanics of prison safety: when officers cannot conduct daily counts or maintain supervision, gangs fill the power vacuum, incarcerated people can unlock their own cells, and classification — the system that separates people by security risk — becomes meaningless because no one enforces the boundaries.

GPS's investigative series Prison Classification Systems & Violence: Misclassification, Overclassification, and Safety Failures mapped how these failures interact. Georgia uses a tool called the Next Generation Assessment (NGA) to assign security levels, but wardens can override the NGA's recommendation. And because GDC policy mandates that sex offenders can never be classified below medium security, independent of their actual institutional behavior, the system systematically overclassifies some populations while underclassifying violent individuals who should be in close custody. The DOJ found that close-security inmates — escape risks, those with assault histories — were being housed in medium-security prisons "not designed or staffed for that population." Research cited by GPS shows that both inmate-on-inmate and inmate-on-officer violence rise when prisons exceed their design capacity, and that classification-driven mismatches directly contributed to deadly riots and murders of staff in other states with analogous breakdowns.

To these staffing and classification failures, GPS's research adds a third, less visible driver: chronic undernutrition. GPS's investigation Slow Starvation in Georgia Prisons: Chronic Undernutrition as Undocumented Cause of Death in GDC Custody documented that GDC SOP 409.04.02 reduces incarcerated people to two meals per day on weekends and holidays — more than 110 days per year — with a maximum 14-hour gap between the evening meal and breakfast. GDC's own food budget, estimated at between $1.60 and $1.69 per person per day, is among the lowest in the country. The investigation reviewed forensic literature showing that chronic undernutrition produces a distinct set of autopsy findings — generalized fat loss, muscle wasting, organ atrophy, gelatinous bone marrow transformation — that routinely go unrecorded in Georgia's death investigations because no protocol requires nutritional status assessment. A national survey by Impact Justice found that 94% of formerly incarcerated people could not eat enough in prison to feel full, 75% were served spoiled food, and over 60% rarely or never had fresh vegetables. GPS's analysis raises the hypothesis that some of the 644 deaths classified as "Natural/Medical" were in fact the end stage of nutritional deprivation, but acknowledges that no peer-reviewed study has yet established this in a U.S. adult prison context.

The Accountability Vacuum

The deaths occur inside a system that, as a matter of deliberate design, lacks external oversight. GPS's legal review Nobody Watches the Watchmen: Independent Correctional Oversight Models and the Statutory Path for Georgia documented that approximately 20 states plus the District of Columbia have an independent prison oversight body; Georgia is not among them. No state-level oversight bill has ever been introduced in the Georgia General Assembly. When the Senate Supporting Safety and Welfare of All Individuals in Department of Corrections Facilities Study Committee (SR 570) convened in 2024, it explicitly declined to recommend an oversight body, with the chair calling the proposal "redundant." A proposal to have the GBI investigate all in-custody deaths was also rejected.

Instead, Georgia's mechanism for post-death accountability is a fractured patchwork. GDC internal investigation reports are classified as state secrets unless the Commissioner declassifies them in writing. The Board of Corrections, whose 19 members are all gubernatorial appointees, conducts no independent inspections and employs no monitors — a captured structure. And while the GBI has superseding authority to autopsy deaths on state property, it is "authorized," not required, to do so, and in practice, the elected coroners in the 155 lay-coroner counties determine whether a death receives close scrutiny. Federal backstops have also failed: the BJA has not published death-in-custody data, the GAO found nearly 1,000 deaths unreported to DOJ in FY2021 alone, and a Senate subcommittee concluded the DOJ's failure to implement the Death in Custody Reporting Act has "deprived Congress and the public of information about who is dying in custody and why."

The litigation record provides a narrow window into what happens when cases do reach court. Law.com reported that a federal judge sanctioned GDC for spoliation of video evidence in a prisoner death case. In another case, correctional officer Angela Butler was reported to have placed a handcuffed prisoner, Hakeem Williams, in a cell with an uncuffed inmate, resulting in a fatal stabbing; the report states that she then lied under oath about violating safety procedures. GPS's own compilation of settlement data, drawn from Georgia DOAS Risk Management records, identified a floor of $50.6 million paid across 261 death-and-injury claims — a figure that, even as a partial accounting, exceeds the annual cost of a fully staffed independent oversight office on any comparator state's model.

Who Is Not Counted as a Victim

Beyond the investigative failures, Georgia's statutory framework structurally excludes incarcerated people from recognition as victims. GPS's research foundation Who Counts as a Victim? Georgia's Statutory Blindness to In-Custody Victimization documented that O.C.G.A. § 17-15-7(c) categorically bars any compensation award "to a victim injured while confined in any federal, state, county, or municipal jail, prison, or other correctional facility." O.C.G.A. § 17-17-3(11) defines "victim" for the Crime Victims' Bill of Rights and expressly excludes any surviving relation who is "in custody for an offense." The Georgia Office of Victim Services, whose mission statement pledges "the highest priority and greatest compassion to those citizens who are most affected by crime, the innocent victims and their survivors," makes no mention of people harmed inside prisons, and the DOJ noted that the Parole Board functions only as a passive "reporting entity for sexual abuse allegations," not as a victim-services provider to incarcerated people.

That exclusion sits in tension with what the research calls the victim-offender overlap: the large body of criminological and public-health evidence, synthesized by GPS, showing that nearly everyone who commits harm has survived it, and that incarcerated populations carry disproportionately high burdens of childhood trauma, sexual abuse, and community violence. Half of women in state prison report prior physical or sexual abuse; 50% of justice-involved youth have four or more adverse childhood experiences. By drawing a bright line around who counts as an "innocent victim," the state's victim apparatus renders invisible the very people whose prior victimization is most strongly associated with later incarceration — and then excludes them again when they are victimized inside.

The Current Trajectory

The deaths have not stopped. GPS-tracked mortality records show that in the most recent weeks, Donald Hugh Young died at Augusta State Medical Prison on August 5, 2026; Steve Allen Burk died at Central State Prison on July 20; and multiple others — at Baldwin, Pulaski, Valdosta, Washington, and beyond — have died in the preceding months, with most causes still listed as unknown by GPS's classification. Earlier in 2026, a fourth inmate death was reported at Hancock State Prison in a single year, and a statewide lockdown followed a gang-related altercation at Smith State Prison. At Washington State Prison, a gang disturbance during visiting hours killed four people. GPS records show that over the past twelve months, 81 distinct sources spread across 10 facilities reported death-in-custody events, with 39 additional sources across 8 facilities reporting that a loved one's life was feared to be in imminent danger.

The federal response, meanwhile, has stalled. GPS's reporting notes that the current administration has halted DOJ civil rights investigations and litigation nationwide, and that 70% of Civil Rights Division attorneys have departed. The DOJ has not yet filed a CRIPA enforcement action against Georgia, and the October 2024 findings letter — however damning — remains a document without a lawsuit behind it. The Senate study committee's recommendations, adopted unanimously in December 2024, focused on facility hardening, mental health services, and federal jamming technology; oversight was left out. The state's budget trajectory adds another layer: GDC's actual spending reached an all-time peak of $1.91 billion in FY2025, a 57% increase in four years, without reversing the death trend.

In Georgia, where the state constitution uniquely contains a Reconstruction-era "Abuse Provision" declaring that "there shall be no abuse of prisoners," the meaning of that clause remains, more than 150 years after its adoption, never fully interpreted by any court. The deaths continue, their causes concealed behind a system of elective office, discretionary autopsy, and sealed records. And the official count — 1,860 dead since 2020 — remains a floor, not a ceiling, on what is known.

Sources: This analysis draws on the October 2024 U.S. Department of Justice findings report, GPS's own investigative coverage across multiple editorial series (Suicide in Georgia State Prisons, Who Decides How They Died, Prison Classification Systems & Violence, Slow Starvation in Georgia Prisons, Nobody Watches the Watchmen, Who Counts as a Victim, and Brown v. Plata: The Legal Blueprint), the GDC SOP 409 series and related internal policies, GPS-tracked mortality data and the associated canonical death count, and reporting from the Atlanta Journal-Constitution, Law.com, and other outlets.