A review project that used artificial intelligence to examine work by economists Emmanuel Saez and Gabriel Zucman says it flagged 2,035 issues across 138 papers, 147 of which could alter headline figures.
The project says 69% of the papers had at least one issue. Among findings it placed in its two higher-severity categories and could classify by direction, it says 97% pointed to claims being overstated. That share applies to this subset, not to all 2,035 findings.
Its website names Claude Opus 5.5 and OpenAI’s GPT-6 Astra as the review systems. The project says a separate adjudicator then challenged flagged issues, and that it recalculated main results using authors’ files and original data where available. The website describes the process but cites no independent review of the complete findings.
One example is a July 2026 working paper co-authored by Ben Tippet and Gabriel Zucman. It proposes a 2% minimum annual tax on total wealth for households with net wealth above £100 million, with personal income taxes credited toward the minimum. The authors estimate £10.41 billion in 2026 revenue after behavioral responses; they separately put administrative costs below 1% of revenue.
The audit calls that estimate a major issue. Using tax-rate figures from research cited in the paper, it calculates revenue of about £8.01 billion. The audit entry says it found no author response to that specific calculation, which is the project’s alternative estimate rather than a revision by the paper’s authors.
The paper argues that the United Kingdom’s super-rich pay lower effective tax rates than ordinary workers and says the proposed levy would reduce the 2026 current-budget deficit. The households above the net-wealth threshold would face the tax. The proposal also includes a 10-year liability after emigration, a period the paper compares with one in U.K. inheritance-tax rules.
The paper’s cover describes it as a discussion document and says it has not undergone peer review or the Observatory’s formal checks for official publications. On Oct. 28, Marginal Revolution blogger Tyler Cowen credited an anonymous reader with the project and called for “a radical revaluation” of the research and related policy arguments.
