The Threat How It Works On-Device vs. Cloud Use Cases Privacy Pricing Developers About Us

AI-Powered Attacks on Mobile Apps

AI leaders are warning that the same tools can be turned to harm. Fraud is moving into mobile apps, driven by new AI bots. Legacy defenses struggle against attacks enabling sophisticated automation at a scale and speed they were never designed to face.

$534B
Lost to Fraud by About 1,200 Companies in One 2025 Survey, 7.7% of Their Revenue (TransUnion)
The loss is only part of the bill: US e-commerce and retail merchants put the total cost at $4.61 for every $1 of fraud loss, counting disruption, investigation, remediation and reputational damage (LexisNexis, 2025).Sources: [1] TransUnion, Oct 2025 · [2] LexisNexis Risk Solutions, Apr 2025
56%
Of Merchants Expect Their Fraud Tool Spending to Rise (MRC, 2026)
Fraud is moving into mobile apps. The industry needs a new approach that can operate where the attacks actually originate - inside the mobile app, on the device itself.Source: [3] Merchant Risk Council, 2026 Global eCommerce Payments and Fraud Report (1,278 merchants)
"AI has fully defeated most of the ways that people authenticate currently, other than passwords."
Sam Altman · CEO, OpenAI, at the Federal Reserve, July 2025 · Transcript
"A bad actor could repurpose those same technologies for a harmful end."
Demis Hassabis · CEO, Google DeepMind, to CNN, June 2025 · Source

The Missing Protection Layer

Today's fraud tools work at the network, in the cloud, or check the device's integrity. Most of them can't see who is physically operating a phone inside a running app, and that's where AI-driven bots now work. Many leading apps already run several fraud and bot SDKs side by side. What's still missing is a continuous human-presence layer on the device.

Why Traditional Defenses Fail

AI bots now operate on real smartphones in phone farms - legitimate OS, real SIMs, no malware. Existing tools can classify them as human and safe. And the next wave is already here: AI agents operating genuine devices end to end.

Endpoint Security

Looks for malware only

  • A clean phone is automatically assumed safe
  • Blind to automation on legitimate hardware
Perimeter Bot Management

Sees requests, not devices

  • Real phones from phone farms look human to perimeter tools
  • Per-request scoring only - not continuous
Cloud Behavioral Analytics

Sends data off the device to produce a score

  • Relies on historical baselines - weak on first sessions or new accounts
  • Privacy and latency tradeoffs
A bot takes over a bank app on a real phoneIllustration
MyBankUser IDPasswordLog injdoe_83••••••••MyBankBalance$12,480.20TransferCardsTransferToAmountSendChoose payeeNew · ACCT 4471$4,900.00Transfer!Held for reviewVerify it is youVerifyBOTBOT SCRIPT00.00launchcom.mybank.app00.50tapuser_id (118, 97)00.80typejdoe_83 · 9 ms per key01.40tappassword (118, 147)01.90type•••••••• · 9 ms per key02.40taplog_in (118, 192)03.40taptransfer (83, 146)04.60tappayee · add ACCT 447106.00tapamount · 4,900.0007.80tapsend (118, 192)CYNOPTA HUMAN SCORE0.620.310.090.030 bothuman 1Motionnone: the phone lies flatChargingon a charger all dayTouchsame pressure, exact centersUsageinstant typing, no pausesApp holds the transferBot detected
Every move is straight and instant, the phone lies flat on a charger, and the score falls with each step. The app holds the transfer before it reaches the bank's backend.
The Next Wave: AI Agents Operating Real Phones

Agents can now drive a real phone from start to finish. Some are sent by your customers, others by attackers, and network, attestation and cloud tools often can't tell them apart. Cynopta shows your app whether a person is behind the phone, so your policy can admit the agents you trust and stop the rest. Read more: Agents Acting for Customers

Cynopta - Inside the phone. Continuous. Scored on the device.

Cynopta detects non-human behavior on clean devices - before a protected action reaches the backend - without presenting any user challenge. Its on-device technology converts raw smartphone signals into a real-time, real-world behavioral profile - enabling a novel way of estimating whether a device is being operated by a human or by automation. Rather than routing device data to the cloud for processing, Cynopta AI models run locally, on-device, delivering instant, continuous bot detection scoring while preserving user privacy.

See Who Is Behind the Phone

Free up to 10,000 assessments a month. Business from $8 a month.