As cybersecurity threats continue to escalate globally, analysts and researchers are working to quantify the financial impact of rising cybercrime costs. The stakes are significant: widely cited industry projections, such as Cybersecurity Ventures' estimate that global cybercrime damages could reach $10.5 trillion annually by 2025, have shaped much of the public and policy conversation — yet such headline figures often lack detailed methodology. Against that backdrop, several recent estimates offer partial but concrete figures, even as the broader public discourse remains heavy on alarm and light on specifics.

One partial estimate of higher cyber costs comes from research aimed at putting numbers on the global annual damage caused by cybercrime. While the number of incidents underlying the estimate is not entirely clear, the figures appear realistic — and arguably modest relative to the level of concern circulating in public discussion. As the author notes, it may not rise to the level of a Gilda Radner "never mind" moment, but it does demonstrate the value of attempting to be concrete.

A separate estimate projects an additional $100 billion in annual cybersecurity costs for the UK alone, representing approximately a twenty percent increase in total UK cyber costs. To put that in the GDP-share framework the author advocates, $100 billion amounts to roughly three percent of UK GDP — a meaningful but not catastrophic scale. That figure also appears plausible to the author.

Brad Carson has been using AI agents to produce cost estimates. While Carson believes the author disagrees with his approach, the author clarifies that he does not — he simply wants exact numbers presented as a share of GDP, which could help bring greater perspective and calm to the conversation.

Superforecaster estimates are also available, offering another structured approach to forecasting cybercrime-related financial losses. These aggregation methods draw on the same discipline used in prediction markets and government foresight programs, applying structured elicitation to reduce bias in forward-looking risk assessments.

Beyond the numbers themselves, the author observes a pattern of errors in financial economics across social media discussions of cybersecurity and AI risk. In particular, he notes that very intelligent people tend to make flawed arguments when the topic arises, especially on the matter of "selling short" — presumably referring to strategies for hedging or profiting from anticipated cyber-driven market disruptions.

The author also addresses what is sometimes called the "beware isolated demands for rigor" response — the argument that requesting rigorous quantification from one side but not others is unfair. He rejects this framing, arguing that those who are genuinely worried about AI should be eager to provide rigor, regardless of whether similar standards are applied elsewhere.

In a more provocative vein, the author proposes a half-serious policy: all expressions of despair, grief, panic, or anger related to AI safety — with or without animated gifs — should be accompanied by a footnote estimating the relevant cost as a percentage of GDP. Such a requirement, he suggests, would inject sanity into the discussion and help drain some of the emotional excess that currently pervades the debate.

Source: Marginal Revolution