TL;DR: An AI model beat a commercial licensing system in under five hours for $3.88, according to Kasada's Q2 2026 threat intelligence report, while 11.7M stolen accounts, 20.9K bot checkouts and 11.3M scraping visits show adversaries are scaling faster and cheaper. The control problem is no longer whether bots exist, but whether fraud and identity defences can keep pace with AI-assisted operator economics.
At a glance
What this is: Kasada's Q2 2026 report argues that AI is reducing the cost of fraud and bot operations while criminal marketplaces, account theft and scraping activity remain high.
Why it matters: For IAM, fraud and digital identity teams, the issue is no longer isolated bot traffic but an industrialised attack economy that can overwhelm account controls, checkout flows and identity verification steps.
By the numbers:
- 11.7M+ stolen accounts were for sale on criminal marketplaces, from 6K sellers.
- 20.9K+ checkouts were completed by bots instead of real shoppers, across 32+ brands.
- 11.3M+ site visits came from just 2 AI companies scraping content, in a single month.
Context
AI has changed fraud operations by lowering the cost of experimentation, scaling and evasion, but it has not changed the underlying business model. Criminals still need stolen credentials, synthetic identities, bot access and monetisation paths, which means the governance problem sits at the intersection of fraud controls, identity assurance and automated abuse detection.
This report is best read as a shift in adversary economics rather than a story about a single new threat. The pressure point for defenders is that AI now accelerates workflows across credential markets, scraping and checkout abuse, which makes static rule sets and slow response cycles less effective.
Key questions
Q: How should security teams respond when AI lowers the cost of fraud operations?
A: They should move from static fraud rules to adaptive controls that combine identity assurance, bot detection and transaction risk. The key is to assume attackers can rent capability cheaply, iterate quickly and shift channels when blocked. That means tighter session checks, stronger recovery controls and faster rule rotation.
Q: Why do stolen accounts and bot abuse need to be governed together?
A: Because they are often different stages of the same abuse chain. A stolen account can provide persistent access, while bot automation turns that access into checkout fraud, scraping or resale value. If teams treat them separately, attackers can move between controls faster than governance can follow.
Q: What are the signs that bot defenses are failing?
A: A common sign is when account creation, login, or recovery traffic looks healthy in aggregate but hidden abuse keeps increasing. Other signals include distorted conversion metrics, repeated support-flow abuse, and a rise in suspicious sessions that pass basic checks but still produce losses or account takeovers.
Q: How can organisations reduce the value of scraped content to attackers?
A: Limit what can be extracted at scale, instrument scraping paths as adversary reconnaissance and connect those signals to account abuse and fraud detection. The goal is not only to block requests, but to make scraped data less useful for targeting, automation tuning and monetisation.
Technical breakdown
How AI changes the economics of bot operations
AI does not create a new adversary class so much as it lowers the friction of existing ones. It helps operators write, test, mutate and rent tooling faster, which reduces the cost of failed attempts and increases the volume of low-quality traffic that still produces revenue. In practice, this compresses the time between tool creation and profitable use, making rate limits, challenge pages and static signatures less durable as standalone controls.
Practical implication: Treat AI-assisted automation as an operating model shift, not a point-in-time spike, and test whether your controls still work when the attacker can iterate continuously.
Why account theft and bot checkout activity belong in the same governance conversation
Stolen accounts and bot-driven purchases are different stages of the same abuse chain. Compromised credentials give operators durable access to customer identities, while automated checkout activity turns that access into inventory drain, fraud or resale value. When the same marketplace can buy, test and operationalise credentials at scale, identity governance has to consider authentication strength, anomaly detection and downstream transaction risk together.
Practical implication: Map credential abuse, account takeover and checkout fraud to a single control path instead of separate teams with separate thresholds.
What scraping tells defenders about exposure and asset value
Large-scale scraping is not just a content issue. It signals that publicly available data can be harvested, repurposed and fed back into fraud workflows, including account targeting, price intelligence and evasion tuning. When scraping volume reaches millions of visits, defenders should assume their content, catalogues and signals are being mined as inputs to broader abuse campaigns rather than viewed in isolation.
Practical implication: Use scraping telemetry as an indicator of adversary reconnaissance and reprice the risk of public content accordingly.
Threat narrative
Attacker objective: The attacker’s objective is to monetise access faster and at lower cost by combining AI-assisted tooling with stolen accounts, bot automation and scraped data.
- Entry begins with rented AI capability and commodity tooling that lower the barrier to trial-and-error abuse across fraud and scraping workflows.
- Credential harvesting and automation then convert stolen accounts, bot sessions and scraped content into repeatable access paths and monetisation channels.
- Escalation follows when operators scale across brands and marketplaces, using the cheapest working method to sustain volume while defenders lag behind.
- Impact is realised through account abuse, checkout fraud, content extraction and the continued resale of stolen identity data.
Breaches seen in the wild
- iOS apps leaking hard-coded secrets: Cybernews found 71% of 156,080 iOS apps leak hard-coded secrets, with open cloud storage and Firebase databases exposing user data.
Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.
NHI Mgmt Group analysis
AI has changed fraud economics before it has changed fraud intent. The article shows that the attacker population has not become more novel, but it has become cheaper to equip and faster to iterate. That matters because most fraud programmes are still structured around volumes, signatures and static risk rules rather than a continuously rented adversary supply chain. The practitioner conclusion is that detection and response now have to assume industrialised iteration, not occasional bursts of abuse.
Credential markets and bot operations are now one governance problem. When stolen accounts, automated checkouts and scraping activity all sit inside the same economic loop, separation between identity teams, fraud teams and application defenders becomes a control gap. This is where threat intelligence helps, not as a feed-only exercise but as a way to connect marketplace signals to observed abuse paths. The practitioner implication is to govern abuse patterns end to end, not as isolated incidents.
Identity assurance must now account for machine-assisted adversaries at scale. The report’s strongest signal is not the headline numbers alone, but the fact that low-cost AI can compress attack preparation and lower the cost of failure. That pushes identity and fraud programmes toward adaptive assurance, stronger session integrity and better linkage between login behaviour and transaction risk. The practitioner conclusion is that trust decisions have to become more dynamic than the attacker’s tooling.
Fraud is increasingly an infrastructure problem, not a point-control problem. The report points to payment rails, marketplaces and automated re-rental as the real leverage points for adversaries. That means teams that only tune front-end defences will keep seeing the same abuse reappear through new infrastructure. The practitioner conclusion is to pursue controls that disrupt reuse, monetisation and operational continuity, not only visible abuse at the edge.
Shadow AI is relevant here because the same automation logic is being adopted by both defenders and attackers. The article shows how AI tools are already being operationalised by criminals as infrastructure, which means unmanaged internal AI use can widen the gap between detection and adaptation. Identity and security teams need governance over how automation is introduced into their own workflows, or they will struggle to match attacker speed. The practitioner conclusion is to govern AI-enabled operations with the same discipline applied to privileged access.
What this signals
Fraud economics are now part of identity governance. When attackers can buy capability, rent automation and reuse stolen credentials cheaply, the control objective shifts from preventing every attempt to reducing the value and lifespan of compromised access. That is especially important for customer identity journeys, where recovery, login and checkout can become a single abuse path.
Adaptive control design matters more than static coverage. The article’s core signal is that adversaries can test many paths cheaply until one works, so defenders need policies that change quickly enough to make that experimentation unprofitable. Agentic AI Security Policy Template is relevant here because the same governance discipline used for AI-enabled workflows can help teams set boundaries for automation, monitoring and retirement.
Marketplace monitoring should feed operational decisions, not just intelligence reports. If 11.7M stolen accounts are circulating and bot economics are this cheap, then the question is not whether attackers exist, but how quickly your controls degrade when they do. Security teams should use marketplace and abuse signals to prioritise which flows need stronger assurance first.
For practitioners
- Tighten step-up checks on high-risk sessions Increase friction when logins come from new devices, impossible travel paths, unusual checkout behaviour or account recovery flows that resemble bot-assisted abuse.
- Correlate credential abuse with downstream transaction risk Join identity telemetry, bot signals and payment or checkout anomalies so compromised accounts can be blocked before they convert into fraud.
- Treat scraping as reconnaissance, not noise Track recurring scraping sources, content extraction patterns and session reuse as indicators that the same data may be feeding wider abuse campaigns.
- Refresh abuse playbooks for AI-assisted iteration Assume attackers can rent, swap and retest tooling quickly, so response playbooks should focus on containment speed, rule rotation and control degradation under load.
Key takeaways
- AI is lowering the cost of fraud, but it is also making existing abuse patterns faster, cheaper and easier to scale.
- The report ties together stolen accounts, bot checkouts and scraping as parts of one operating model rather than separate problems.
- Defenders need controls that adapt quickly enough to reduce the value of compromised access and the profitability of automation.
Related NHI Mgmt Group resources: Threat Intelligence Guide · Threat Intelligence Platform Buyer's Guide · Email Security Buyer's Guide · Email Security Guide
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
MITRE ATT&CK and OWASP API Security Top 10 address the attack and risk surface, while NIST CSF 2.0, CIS Controls v8 and NIST SP 800-53 Rev 5 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| MITRE ATT&CK | TA0006; TA0009; TA0010 — Credential Access; Collection; Exfiltration | The report centres on stolen accounts, scraping and monetisation paths that map to ATT&CK abuse stages. |
| Recommendation — Map bot and credential abuse to ATT&CK stages and use them to prioritise detection and containment. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The article is about controlling misuse of accounts, sessions and access paths across digital journeys. |
| Recommendation — Apply PR.AA-05 to limit abusive access paths and tighten authorisation around sensitive flows. | ||
| CIS Controls v8 | CIS-5 — Account Management | Stolen accounts and reused credentials are central to the abuse patterns described in the report. |
| Recommendation — Use CIS-5 to identify, govern and remove account paths that enable repeat abuse. | ||
| NIST SP 800-53 Rev 5 | IA-5 — Authenticator Management | Credential compromise and reuse are core to the threat pattern Kasada describes. |
| Recommendation — Apply IA-5 to harden authenticator lifecycle controls and reduce the value of stolen credentials. | ||
| OWASP API Security Top 10 | API5 — Broken Function Level Authorization | The article touches checkout and automation abuse where authorisation gaps can amplify fraud. |
| Recommendation — Audit sensitive flows for function-level authorisation gaps that bots can exploit at scale. | ||
Key terms
- Bot Fraud: Bot fraud is the use of automated scripts or synthetic accounts to imitate legitimate users and exploit digital services. In onboarding and account-opening flows, it typically aims to extract incentives, distort metrics, or bypass identity checks at scale.
- Credential Market: Credential market is the trade in stolen usernames, passwords, session tokens, and other authentication data. Attackers buy credentials to take over accounts, bypass defences, and launch fraud or ransomware, so defenders should enforce strong MFA, monitor for reuse and leakage, and investigate impossible logins or unusual token use.
- Adaptive Assurance: A governance approach that adjusts identity checks, session scrutiny and transaction controls based on risk signals. It matters when attackers can change tactics quickly, because static controls tend to lag behind automated abuse and rented capability.
- Scraping Reconnaissance: The use of automated collection to map content, pricing, flows or defensive behaviour before a broader abuse campaign. In practice, scraping is often an intelligence-gathering stage that helps attackers tune later fraud or credential attacks.
Deepen your knowledge
NHI Foundation Level course, the industry's only accredited NHI security programme, covers NHI governance, identity lifecycle and machine identity security. It helps practitioners connect identity controls to fraud, access abuse and the broader security programme.
Published by the NHIMG editorial team on October 22, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org