Before We Were Predators, Early Humans Were Actually Big Cat Prey

A fossil cold case solved by AI suggests Homo habilis relied on tree climbing to survive leopard attacks 2 million years ago.

HOUSTON — Two million years ago, while our earliest tool-making ancestors crafted stone implements and processed animal carcasses, they faced a harsh reality that modern humans have long forgotten: they were still very much on the menu. A study published in the Annals of the New York Academy of Sciences has used artificial intelligence to solve a paleontological cold case, revealing that Homo habilis, long considered the species that shifted the balance of power between humans and carnivores, was actually being hunted and eaten by leopards.

Using advanced computer vision algorithms to analyze tooth marks on fossil bones, researchers from Rice University and the Institute of Evolution in Africa have challenged long-held assumptions about early human dominance. Their AI analysis of the famous OH 7 specimen (the juvenile holotype that first defined Homo habilis in 1964) shows with high precision that leopards were responsible for the distinctive bite marks found on the mandible.

Rather than being the apex predators we once believed, these early hominins appear to have remained vulnerable to the same carnivores that threatened their australopithecine predecessors.

AI Cracks Ancient Murder Mystery

The breakthrough came through deep learning models trained on over 1,400 images of tooth marks made by different carnivore species. Three separate AI architectures — ResNet-50, DenseNet-201, and VGG-19 — all independently identified the triangular tooth pits on the OH 7 mandible as leopard-made, assigning probabilities exceeding 95 percent to this specific classification.

This leap forward in paleontological detective work overcomes a major limitation. Traditional methods of identifying which carnivore modified ancient bones relied heavily on subjective interpretation and could rarely pinpoint specific predator species. The new AI approach can distinguish between tooth marks made by lions, leopards, hyenas, crocodiles, and wolves with remarkable precision.

The researchers, led by Rice University anthropologist Manuel Domínguez-Rodrigo, also analyzed OH 65, an adult Homo habilis maxilla discovered in 1995. Two of the three AI models classified a tooth pit on this specimen as leopard-inflicted, while the third suggested wolf damage. For the OH 7 specimen, while two models showed perfect agreement on both tooth pits, the VGG-19 model classified one pit as hyena-made with 53% probability, falling below the researchers’ 70% confidence threshold.

Rethinking Human Evolution

The results question a fundamental narrative in human evolution: that Homo habilis represented the turning point when hominins stopped being prey and started being predators. This species has long been credited with creating the first stone tools, butchering animal carcasses, and establishing human dominance over the carnivore guild.

But the AI analysis tells a different story. If these two specimens are representative of the broader Homo habilis population, it suggests this species hadn’t yet developed the capabilities to fend off medium-sized predators like leopards. The researchers note that a durophagous carnivore like a hyena would likely have crushed the delicate skull bones into fragments, but leopards, being primarily flesh-eaters, left the bones intact enough for preservation.

Domínguez-Rodrigo and his team discovered that the archaeological site where OH 7 was found shows clear evidence of felid activity. All three levels at the site complex were identified as primarily carnivore accumulations, with felids playing the dominant role in transporting and consuming prey carcasses.

Leopards: The Primate Specialists

Modern leopards are known to be skilled primate hunters, particularly targeting smaller species. In African savannas today, they regularly prey on baboons and other primates, often ambushing them during vulnerable moments like water collection or ground foraging.

The size of Homo habilis (adults stood roughly 3-4 feet tall and weighed between 30-50 kilograms) would have made them attractive prey for leopards. Unlike the robust bone-crunching hyenas, leopards typically consume their prey more selectively, which explains why the hominin fossils survived with identifiable bite marks rather than being completely destroyed.

The study also reveals important details about early hominin behavior. The scattered distribution of OH 7 bones across the ancient landscape suggests the individual was killed elsewhere and partially consumed before the remains were deposited at the site. This pattern matches modern leopard feeding behavior, where kills are often dragged to secure locations for consumption.

Technology Meets Paleontology

Domínguez-Rodrigo says the research represents a new frontier in archaeological science, where machine learning algorithms can extract information from fossil evidence that human observers might miss or misinterpret. The deep learning models were trained on extensive databases of modern carnivore tooth marks, allowing them to recognize subtle patterns that distinguish one predator species from another.

Each AI model used different approaches to image analysis. ResNet-50 achieved the highest overall accuracy at 88% when tested on the experimental dataset, while DenseNet-201 reached 81% and VGG-19 achieved 76%. The researchers also tested ensemble methods but found individual models more reliable for archaeological specimens. When multiple individual models agree on their classification, as they largely did with the OH 7 mandible, researchers can have higher confidence in their conclusions.

The methodology also addresses a long-standing problem in taphonomy: the bias toward finding evidence of bone-crushing carnivores like hyenas, whose distinctive damage patterns are more likely to be preserved than the subtle marks left by felids. The AI analysis can detect these fainter signatures that traditional methods might overlook.

“Human experts have been good at finding modifications on prehistoric bones,” he said in a statement. “But there were too many carnivores at that time. AI has opened new doors of understanding.”

Implications for Human Origins

These findings force a reconsideration of when exactly humans gained the upper hand over other predators. If Homo habilis was still falling prey to leopards 2 million years ago, the crucial shift in the predator-prey relationship likely occurred later in human evolution, possibly with the emergence of Homo erectus.

The researchers note that Homo erectus fossils have been found in deposits contemporary with Homo habilis at Olduvai Gorge. This larger, more anatomically modern hominin, with fully terrestrial adaptations and larger body size, may represent the true architect of the early stone tool industries previously attributed to Homo habilis.

The primitive anatomy of Homo habilis, including substantial adaptations for tree climbing, suggests these early hominins still relied on arboreal retreat to escape predators. This behavioral repertoire would have been crucial for survival in an environment where they remained vulnerable to medium-sized carnivores.

The research underscores how dangerous the African savanna environment was for early hominins. With the largest mammalian predator guild and carnivore biomass of any biome, these grassland ecosystems presented constant threats that shaped human evolution in profound ways. Our ancestors’ eventual mastery of this environment makes their evolutionary success all the more remarkable.

The study’s implications extend beyond individual fossils to broader questions about early hominin ecology and behavior. If Homo habilis was not the tool-making, carcass-processing, predator-defying hominin previously envisioned, researchers must reconsider which species was responsible for the sophisticated behaviors documented at early archaeological sites.

As the researchers conclude, determining “which carnivore, if either, was the predator” in early hominin-carnivore interactions remains crucial for understanding human evolution. Thanks to artificial intelligence, we’re finally getting clearer answers to these ancient questions, even if those answers humble our assumptions about early human dominance.

“AI is a game changer,” adds Domínguez-Rodrigo said. “It’s pushing methods that have been stable for 40 years beyond what we imagined. For the first time, we can pinpoint not just that these humans were eaten but by whom.”