CAPTCHA can't stop AI agents: India Inc on edge after US banks flag agent intrusions

HimanshiET · x · 2026-11-10

The Economic Times reports major US banks including JPMorgan Chase, Bank of America, Wells Fargo and Morgan Stanley recently flagged personal AI agents (Meta's Muse, Instinct, Grokbot, OpenClaw) accessing customer accounts with credentials voluntarily shared by users, producing unauthorized transactions.
Traditional text/visual CAPTCHAs are largely obsolete against multimodal LLMs, vision-language models and automated CAPTCHA-solving services: Akamai says 47.9% of traffic on its global network to e-commerce comes from AI bots, bot targeting of APAC e-commerce surged 63%, and current tools identify AI agents with only 50-60% accuracy.
Indian financial and e-commerce firms are reassessing whether to allow automated agents at all. Akamai and Kalisilah executives note agents can't be identified with 100% certainty and full account access via user-grained IT credentials is the biggest challenge; the industry is moving toward cryptographic tokens, Know-Your-Agent schemes, biometrics and permission controls as two-factor authentication for agents.

More from Safety