Mollick: Serious AI incidents surprisingly rare at a billion users, while positive uses are underrated

Wharton professor Ethan Mollick posted on October 6–7 with a counterintuitive observation: with AI already having roughly a billion users and hundreds of millions of enterprise users, genuinely severe AI incidents over the past few years have been relatively rare. He stressed this doesn't mean it will stay that way, but at current scale of use, the low incident rate is genuinely surprising to long-time AI watchers.

Confirmed

Mollick explicitly stated: catastrophic AI incidents are rare at a billion-user scale; this judgment isn't guaranteed to persist, but it is unexpected (m1, m2)
He added: AI's positive use cases are similarly underrated — existing evidence suggests AI's value in providing health information, financial advice, and supporting entrepreneurship is systematically underestimated and under-discussed (m2, m5)
In response to skepticism, Mollick noted that roughly 15% of the global population already uses these systems in their personal and business lives, and adoption so far has gone far more smoothly than most expected (m3)

Not Yet Confirmed

User @iruletheworldmo pressed: how many of these users actually use models with such capabilities, and how many have such needs — this challenge received no data-backed response (m4)
Mollick's claim that "positive use cases are underrated" is based on his synthesis of existing evidence; the post materials cite no specific research data

Why It Matters

The observation offers a scale-based perspective for AI safety discussions: the gap between adoption speed and incident rates could reshape public and regulatory perceptions of AI risk
Reposter @dreamwieber added: tragic events easily make the news, but how many times AI has prevented users' harmful actions is unknowable — suggesting public opinion may carry a survivorship-bias-style tilt toward negative narratives (m5)

2026-10-06 ~ 2026-10-07 · 5 related posts

Primary sources

Ethan Mollick: surprisingly few terrible AI incidents across a billion users — emollick · 2026-10-06 23:14 (UTC+08:00)
Ethan Mollick: 15% of the World Already Uses AI Daily, and It's Calmer Than Expected — emollick · 2026-10-07 00:47 (UTC+08:00)
Ethan Mollick: with a billion users, catastrophic AI incidents remain surprisingly rare — emollick · 2026-10-07 03:12 (UTC+08:00)

[source] Ethan Mollick: surprisingly few terrible AI incidents across a billion users — emollick · 2026-10-06
[source] Ethan Mollick: 15% of the World Already Uses AI Daily, and It's Calmer Than Expected — emollick · 2026-10-07
Mollick: AI's positive use cases in health and financial advice are undermeasured — dreamwieber · 2026-10-07

2 near-duplicate retellings: iruletheworldmo · emollick