{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,14]],"date-time":"2026-08-14T14:54:53Z","timestamp":1786719293391,"version":"build-2736575974"},"posted":{"date-parts":[[2026]]},"group-title":"SSRN","reference-count":0,"publisher":"Elsevier BV","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"abstract":"CAPTCHA systems were designed around the premise that visual challenges are cheap for humans and expensive for machines. The arrival of frontier multimodal AI in 2023-2024 has been argued to invert that asymmetry. I test for the inversion across three publicly observable channels and find that the posted-price evidence is the most informative.\n<div>\n The central empirical contribution is a longitudinal panel of 1,005 posted-price observations from Wayback Machine snapshots of 2Captcha and Anti-Captcha across 37 monthly samples, January 2023 to December 2025, covering 56 distinct (service \u00d7 captcha-type) panels. After normalising for currency switches in the cached pages, I detect 11 repricing episodes at 2Captcha and zero at Anti-Captcha. Two episodes fall on a single day in September 2023; nine fall in two adjacent snapshots in May and June 2025. All eleven are cuts from $2.99 to $1.45 per thousand solves \u2014 discrete jumps rather than calibrated percentage moves. The posted reCAPTCHA v2 bypass price is constant at $1.00\u2013$2.99 per thousand solves across thirty-six months; a permutation test confirms a single break in the 2Captcha Turnstile series at the September 24, 2023 snapshot. Within-wave dependence (the seven June 2025 cuts most likely reflect a single repricing decision, not seven) means that the eleven price changes should not be treated as eleven independent draws around the seven dated AI events. Descriptively, one repricing decision sits four days before Cloudflare's Turnstile announcement, and a second, much larger decision sits twelve to thirteen months after the last AI event in the panel.\n<\/div>\n<div>\n Two corroborating channels triangulate the posted-price finding. Cumulative abnormal returns of four publicly listed bot-defence vendors (Cloudflare, Akamai, Fastly, F5) around seven AIcapability events are not statistically distinguishable from those of a matched cybersecurity placebo (Qualys, Zscaler): the mean treated\u2013placebo difference is +1.67 percentage points (bootstrap SE 2.43, 28 firm-event observations), in the opposite direction to the asymmetric-costcollapse prediction. The panel rules out effects of approximately eight percentage points or larger. Every Cloudflare 10-Q and 10-K filing in the window contains rising aggregate AI mentions and generic AI-cyber-risk language, but the specific AI-bot, AI-crawler, and AI-scraper vocabulary that Cloudflare employs extensively in its blog and press communications during the same period does not appear in any of the sixteen filings.&nbsp;\n<\/div>\n<div>\n The 2023-09-24 cut at 2Captcha precedes Cloudflare's Turnstile general-availability announcement (September 28, 2023) by four days; the temporal evidence on Wave 1 cannot identify a direction of causation. The May\u2013June 2025 wave is genuinely ambiguous: it post-dates the last AI capability event in the panel by twelve to thirteen months, too long to support a fastpass-through reading but potentially consistent with a delayed cost-pass-through on a multiquarter lag. I treat the 2025 wave as the central interpretive challenge of the paper.&nbsp;\n<\/div>\n<div>\n I observe posted rack rates, not transacted prices. If effective prices fell through unobserved volume discounts, off-platform negotiation, or Telegram-channel relationship pricing, the present method would not detect the change. The joint evidence on publicly observable price signals points to delayed and partial pass-through into the specialty-captcha tier during 2025, while posted prices for the major high-volume captchas (reCAPTCHA, hCaptcha, GeeTest, FunCaptcha-high) remained rigid through the 2023\u20132024 window of frontier capability releases.&nbsp;\n<\/div><\/jats:p>","DOI":"10.2139\/ssrn.6760919","type":"posted-content","created":{"date-parts":[[2026,5,20]],"date-time":"2026-05-20T08:41:47Z","timestamp":1779266507000},"source":"Crossref","is-referenced-by-count":0,"title":["Pricing the Collapse of Proof-of-Human: Posted-Price, Equity, and Disclosure Evidence on AI-Capability Events, 2023-2025"],"prefix":"10.2139","author":[{"ORCID":"https:\/\/orcid.org\/0009-0005-8477-9393","authenticated-orcid":true,"given":"Boon Chuan","family":"Lim","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","container-title":[],"original-title":[],"deposited":{"date-parts":[[2026,8,14]],"date-time":"2026-08-14T13:47:53Z","timestamp":1786715273000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.ssrn.com\/abstract=6760919"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026]]},"references-count":0,"URL":"https:\/\/doi.org\/10.2139\/ssrn.6760919","relation":{},"subject":[],"published":{"date-parts":[[2026]]},"subtype":"preprint"}}