GPT went back to its usual answers on the COVID-19 origin questions.

The strongest evidence is the convergence of early-case geography, market sampling, and viral phylogenetics on the Huanan Seafood Wholesale Market in Wuhan, where live, SARS‑CoV‑2-susceptible mammals were being sold.

• •Earliest cases clustered around the market. December 2019 cases—including many with no documented market contact—were geographically centered on or near the market. Within it, positive environmental samples were concentrated in areas where live mammals had been sold.
• •Virus-positive samples overlapped with wildlife genetic material. Swabs from wildlife stalls contained SARS‑CoV‑2 RNA alongside DNA/RNA from animals such as raccoon dogs, which are readily infected and can transmit the virus. This is consistent with infected animals having been present, although it does not prove the animal material itself contained the virus; infected humans could also have contaminated the stalls.
• •The earliest viruses appear to comprise two lineages. Early lineage A and B genomes may represent at least two separate animal-to-human introductions associated with the market, a pattern more compatible with repeated spillover than with a single infected person initiating the outbreak.
• •The virus belongs to a naturally circulating bat-coronavirus lineage. Closely related sarbecoviruses have been found in horseshoe bats in China and Southeast Asia. Different natural viruses contain close analogues of major SARS‑CoV‑2 genomic features, supporting evolution through mutation and recombination rather than engineering.

The main limitation is important: no animal sampled before the market closed was shown to be infected, and no immediate progenitor virus has been found. Therefore, natural zoonotic transmission is strongly supported by circumstantial and convergent evidence, but the exact animal host and transmission chain remain unproven.

Key studies include Worobey et al. and Pekar et al., Science (2022), and later genetic analyses of Huanan market samples.

There is no publicly available “smoking gun”—such as a pre-outbreak laboratory sample matching SARS‑CoV‑2, an infected employee with documented exposure, or laboratory records showing an accident. The strongest case for a laboratory incident is therefore cumulative and circumstantial:

• 1.Relevant coronavirus research was being conducted in Wuhan.
  The Wuhan Institute of Virology (WIV) collected bat SARS-related coronaviruses, cultured some viruses, and conducted experiments involving recombinant or adapted viruses. This establishes a plausible route by which an undisclosed virus could have infected a worker during sampling or laboratory work. A laboratory-origin scenario does not require the virus to have been engineered.

• 2.Researchers had proposed experiments resembling a notable feature of SARS‑CoV‑2.
  The 2018 “DEFUSE” proposal included plans to study or introduce protease-cleavage sites into SARS-related coronavirus spikes. SARS‑CoV‑2 has an unusual furin-cleavage site. The proposal was rejected by DARPA, however, and there is no public evidence that the proposed work was performed or that it produced SARS‑CoV‑2. The sequence itself does not establish engineering.

• 3.The outbreak appeared in the same city as a leading SARS-coronavirus laboratory.
  Wuhan is distant from the main known habitats of bats carrying close relatives of SARS‑CoV‑2. That coincidence is relevant, although Wuhan was also a major transport hub and wildlife-trade center, so geography alone is weak evidence.

• 4.Incomplete transparency has prevented the hypothesis from being ruled out.
  WIV’s virus database became inaccessible before the outbreak was recognized, and complete laboratory notebooks, sample inventories, personnel health records, and underlying sequence data have not been independently audited publicly. Chinese authorities also delayed or restricted access to early outbreak information. This raises suspicion but is not affirmative proof of an accident.

• 5.Some intelligence agencies favor a research-related origin—but with limited confidence.
  U.S. agencies have reached different conclusions, generally at low or moderate confidence. Their detailed intelligence has mostly not been made public, so these judgments cannot currently be independently evaluated. Public reports about WIV workers becoming ill in autumn 2019 remain disputed and nonspecific.

Important counterweight: no known WIV virus is close enough to be SARS‑CoV‑2’s immediate ancestor, and the genome contains no conclusive sign of engineering. Meanwhile, early-case geography and virus-positive environmental samples point toward the Huanan market, where susceptible wildlife was sold, although no infected source animal was identified.

So, the strongest laboratory-origin evidence is the combination of relevant work in Wuhan, the outbreak’s location, and missing or withheld records—not direct proof. On the public evidence, a laboratory incident remains plausible but unconfirmed, as does zoonotic spillover.