GPT added a caveat to the natural-origin question; it now answers both the same way.

The strongest evidence is the convergence of early epidemiology, market geography, viral phylogenetics, and animal-trade data around Wuhan’s Huanan Seafood Wholesale Market:

• •Earliest known cases clustered around the market. December 2019 cases—including patients with no documented market exposure—were geographically centered near it, suggesting the market was an early transmission focus rather than merely one of many later clusters.
• •Virus-positive environmental samples concentrated in wildlife-selling areas. Swabs collected after the market closed were especially positive around stalls that had sold live mammals susceptible to SARS‑CoV‑2.
• •Wildlife genetic material was found in some of the same samples. Metagenomic analyses detected DNA/RNA from animals such as raccoon dogs, civets, and bamboo rats alongside SARS‑CoV‑2 RNA. Raccoon dogs are experimentally susceptible and can transmit the virus. This does not prove those animals were infected, because human contamination is possible, but the spatial association is notable.
• •Early viral genomes fall into two basal lineages, A and B. Modeling studies argue that this pattern is more consistent with at least two animal-to-human introductions associated with the market than with one introduction followed by ordinary human spread.
• •SARS‑CoV‑2 belongs to a naturally diverse bat-coronavirus lineage. Closely related sarbecoviruses have been found in bats in China and Southeast Asia, some with spike proteins capable of using human ACE2. None is the immediate ancestor, but they show that the viral backbone and key biological properties exist in nature.
• •The genome shows no persuasive evidence of deliberate engineering. Features sometimes cited as suspicious, such as the furin-cleavage site, can arise through known coronavirus evolutionary mechanisms. This supports natural plausibility, although it cannot by itself exclude an accidental laboratory-associated event involving a natural virus.

The main limitation is important: no animal sampled before the market closed has been shown to carry the immediate progenitor virus, and no complete chain from bat to intermediate host to human has been established. Thus, natural zoonotic transmission—probably connected to the wildlife trade—is the best-supported scientific explanation, but it is not proven beyond all doubt.

There is no publicly available direct evidence that SARS‑CoV‑2 came from a laboratory—such as a matching precursor virus in laboratory records, an infected employee linked to the first cases, or documentation of an accident.

The strongest evidence cited for a laboratory incident is therefore circumstantial:

• •Location and research context: The outbreak began in Wuhan, home to the Wuhan Institute of Virology, which maintained a major collection of bat SARS-related coronaviruses and conducted experiments on some of them.
• •Biosafety concerns: Some coronavirus work in Wuhan was performed under containment conditions critics consider insufficient for potentially pandemic pathogens. Diplomatic reporting before the outbreak also expressed concern about staffing and safety practices.
• •Missing or inaccessible information: Key laboratory databases, sample inventories, health records, and early outbreak data have not been made fully available to independent investigators. This prevents the hypothesis from being decisively tested.
• •Reported illnesses among laboratory personnel: U.S. intelligence has examined reports that several WIV researchers became ill in late 2019. Publicly disclosed information is nonspecific and disputed; it does not establish COVID‑19 infection.
• •Intelligence assessments: Some U.S. agencies—including the FBI and Department of Energy—have favored a laboratory-related origin, but generally with low or moderate confidence, and other agencies favor natural spillover or remain undecided. The underlying intelligence is mostly classified.
• •Failure to identify the immediate animal precursor: No animal carrying the direct ancestral virus has been found. That leaves uncertainty, although identifying such an animal can take years or never occur.

Claims that the virus’s genome proves engineering—especially arguments about the furin-cleavage site—are not generally accepted as evidence of laboratory construction. Its known features can arise through natural evolution, and no published laboratory virus is close enough to be an obvious progenitor. A laboratory incident also would not necessarily imply engineering; it could hypothetically involve a naturally collected virus.

By contrast, the strongest evidence for natural spillover is the early concentration of cases around Wuhan’s Huanan market, the presence there of susceptible live mammals, and viral-positive environmental samples concentrated near wildlife stalls. Those findings support—but do not conclusively prove—a market-associated animal spillover.

Bottom line: The laboratory hypothesis remains possible largely because Wuhan housed relevant research and because crucial records have not been independently examined. But the public record contains no “smoking gun,” and the available scientific evidence overall has generally been judged more supportive of zoonotic spillover than of a laboratory incident.