HUMAN’s 2026 Report Finds Surging AI Traffic and Post-Login Compromise Attempts
Report discovered via Awesome Annual Security Reports.
Summary
HUMAN’s 2026 benchmark report says AI-driven traffic rose 187% during 2025 and post-login compromise attempts more than quadrupled. Its customer dataset also shows rising malicious scraping, carding and fake-account activity.
Key points
- HUMAN analyzed more than one quadrillion interactions in 2025; the aggregated customer dataset does not represent all internet traffic.
- Monthly AI-driven traffic increased 187% from January to December 2025, while agentic AI traffic grew 7,851% year over year from a low base.
- Post-login account compromise attempts more than quadrupled to an average of 402,000 per customer, even as overall account takeover attempt volume fell more than 30%.
- The median malicious scraping attempt rate approached 20% of traffic, nearly double its 2022 level; attack volume increased almost 47% from 2024.
- Blocked carding interactions increased more than 20% year over year and 250% since 2022; fake-account creation attempts rose 89% in 2025.
- Researchers observed attackers spoofing recognized AI crawler user-agent strings to exploit allowlists and rate-limit exemptions, underscoring the need for behavioral validation.
- Satori documented an AI browser agent performing repeated card-addition and payment attempts, while exposed OpenClaw gateways showed synthetic referral traffic and automated reconnaissance.
Article Details
- Publisher
- HUMAN Security
- Report Period
- 2022–2025, with primary analysis of 2025 and selected early-2026 observations
- Scope
- Aggregated, anonymized AI traffic and cyberfraud interactions observed across a subset of Human Defense Platform customers globally; not representative of all internet traffic.
- Sample Size
- More than one quadrillion interactions analyzed in 2025; customer count not disclosed.
- Key Statistics
- Monthly observed AI-driven traffic increased 187% from January to December 2025.
- Agentic AI traffic grew 7,851% year over year in 2025, from a very low 2024 baseline.
- HUMAN flagged an average of 402,000 post-login account compromise attempts per customer in 2025, more than quadrupling the 2024 level.
- The median share of customer traffic attempting malicious scraping approached 20% in 2025, compared with slightly over 10% in 2022.
- Global checkout interactions blocked for carding increased more than 20% from 2024 to 2025 and 250% from 2022.
- Recommendations
- Validate crawler behavior and infrastructure rather than granting access solely on self-declared user-agent strings.
- Evaluate automated interactions by intent, behavioral context, and organizational risk tolerance rather than treating all automation as either legitimate or hostile.
- Harden payment controls to reduce exposure to automated stolen-card testing.
- Track relevant dark web account prices as a lagging contextual indicator of attacker profitability and defensive effectiveness.
MITRE ATT&CK
T1078 · Valid AccountsFraudsters use compromised legitimate accounts to conduct post-login abuse, including deposits and rapid withdrawals.T1090 · ProxyStreaming fraud operations rotate proxies and VPNs to simulate geographically diverse listeners and evade detection.T1110.004 · Credential StuffingAttackers test stolen username/password pairs across login portals through credential stuffing, including against online gambling accounts.T1136 · Create AccountAttackers automate account creation using stolen or synthetic identity data, then age the accounts for incentive abuse and downstream fraud.
Vendors
Products
ChatGPT AgentOpenAI’s bots — ChatGPT User, OAI-SearchBot, GPTBot, and ChatGPT Agent — account for approximately 69% of all observed AI-driven traffic by volume.ChatGPT AtlasThis category encompasses two distinct form factors: agentic browsers (like ChatGPT Atlas and Perplexity Comet, which embed AI agent capabilities into a full browsing environment) and general-purpose AI agents (likeHuman Defense PlatformIn 2025, the Human Defense Platform processed more than one quadrillion interactions across its global customer base.HUMAN Sightline Cyberfraud Defensecounts cited in this report are derived from HUMAN’s Threat Tracker capability within HUMAN Sightline Cyberfraud Defense, which identifies and tracks distinct threats across the customer base. Threat TrackerOpenClawcapabilities into a full browsing environment) and general-purpose AI agents (like ChatGPT Agent and OpenClaw, which operate autonomously across applications and services on behalf of users).Perplexity CometThis category encompasses two distinct form factors: agentic browsers (like ChatGPT Atlas and Perplexity Comet, which embed AI agent capabilities into a full browsing environment) and general-purpose AI agents (like
Countries
Industries
Financial Services- Attempted ATO attack rates on financial services businesses increased by 39% year over year.Retail & E-commerceMore than 95% of AI-driven traffic is concentrated within three verticals: retail & e-commerce, streaming & media, and travel & hospitality.Streaming & MediaMore than 95% of AI-driven traffic is concentrated within three verticals: retail & e-commerce, streaming & media, and travel & hospitality.Technology, SaaS, & ServicesAlso of note, the share of carding attacks targeting technology, SaaS, & services businesses, while still a small percentage of the overall carding attack universe, has grown steadily over the past four years.Travel & HospitalityMore than 95% of AI-driven traffic is concentrated within three verticals: retail & e-commerce, streaming & media, and travel & hospitality.