AI Call Defence Hits 1 Billion Flagged Calls: What Customer Support Teams Need to Do Now
Virgin Media O2, working with Hiya, has now labeled more than 1 billion suspected scam and spam calls for O2 customers. The free Call Defence service uses AI to analyse unknown calls in real time and warns customers on-screen before they pick up. Alongside call labels, O2 has also blocked more than 1 billion scam texts.
About 70 million calls each month are now flagged as suspected scam or spam. Calls labeled "suspected scam" are answered 42% less often and last 89% less time than unflagged calls-clear proof that real-time warnings change customer behavior and cut exposure.
How Call Defence Works (and why it matters to your team)
When an unknown number rings, Call Defence uses Adaptive AI to evaluate behavior and multiple data points in real time. If the call looks risky, it displays a clear label on the customer's screen. Known fraudulent calls are also blocked outright.
For support leaders, this means fewer high-risk conversations reaching agents, faster dismissals of scam attempts, and less social engineering pressure on your teams. But it also changes what customers expect from you on identity checks and callback flows.
Top Scam Themes in January 2026
- Amazon: Fake support about a disputed order or fraudulent purchase; aims to steal account credentials or card details.
- HMRC: Live or robocalls threatening legal action for unpaid taxes.
- Banking: Impersonation of fraud, security, or customer service teams to gain account access or prompt "safe" transfers.
- Payment services - VISA: Claims of a fraudulent charge or payment issue; pushes victims to share card info-especially common during busy shopping periods.
- Insurance: Impersonation of insurers about policy issues, missed payments, or urgent claims to pressure for personal or payment data.
Agent Playbook: Handling Flagged or Suspicious Calls
- Never authenticate the caller with sensitive info. Agents should not ask for full passwords, one-time passcodes, or complete card numbers over the phone.
- Use outbound callbacks from verified numbers only. If a case is sensitive, end the call and call back using a number listed on your official site or app.
- Route "account access" and "urgent payment" requests carefully. Escalate to your fraud or security workflow; follow strict verification steps.
- Tag, record, and share patterns. Create dispositions for "suspected scam," note spoofed brands, and share weekly trend summaries with your fraud team.
- Coach for calm, concise responses. Short, firm scripts reduce social engineering risk and AHT on suspected scam calls.
Scripts You Can Hand Agents Today
- If a customer reports a flagged call: "Thanks for checking. For your safety, we'll never ask for full passwords or one-time codes. If you're unsure, hang up and call us using the number on our website or app."
- If the caller pressures for urgent payment: "I can't process that request on this call. I'll end the call now and you can reach us on the official number for verification."
- If brand impersonation is suspected: "We're aware of impersonation attempts. I'll log this and share guidance so you can secure your account."
Reporting Matters: Use 7726
Virgin Media O2 urges everyone to report suspicious calls and messages to 7726 (it's free). These reports help investigate and block numbers used by fraudsters and improve future detection.
For quick guidance on reporting phishing texts and calls, see the UK's official advice from the National Cyber Security Centre: Report suspicious emails, texts and websites.
What to Update This Week
- Knowledge base: Add Call Defence FAQs, 7726 steps, and the current top scam themes.
- IVR and on-hold: Briefly mention that scam warnings may appear and remind customers to use official callback numbers.
- QA checklists: Add items for verification hygiene, refusal of sensitive data, and correct scam-tagging.
- Fraud escalations: Define a single intake channel and SLA for suspected impersonation or account-access attempts.
Metrics to Watch
- Flagged-call answer rate and duration: Track reduction against your baseline and share with fraud teams.
- Volume of 7726 reports from your customers: Measure week over week; run micro-campaigns if reports drop.
- Impersonation cases by brand: Prioritize playbooks for the top three targeted brands in your vertical.
- Agent adherence: Monitor no-OTP policy, verified callback usage, and correct dispositioning.
What Leaders Are Saying
"At Virgin Media O2, customers are our top priority, and we're investing heavily to keep them safe from scams and ensure we're a network they can rely on. This includes using the latest AI technology for good, to proactively stop fraudsters from getting to our customers.
Having warned our customers about more than 1 billion calls, and prevented more than 1 billion dodgy messages from ever reaching them, today marks another milestone in our fight against fraudsters. With scammers stopping at nothing - impersonating well-known brands and government departments - to earn victims' trust, it's never been more important to stay vigilant and report suspicious activity to 7726." - Murray Mackenzie, Director of Fraud Prevention at Virgin Media O2
"We've all experienced that moment of doubt when an unknown number appears on our screen. A phone call should create connection, not anxiety. By working with Virgin Media O2 to label 1 billion suspected scam calls in real time, we're giving customers clarity before they ever pick up. When customers are warned in real time, scammers lose their advantage." - Alex Algard, CEO and Founder of Hiya
Why This Changes Your Support Strategy
Real-time call labels shift customer behavior before your team says a word. Your job is to tighten verification, reduce risky conversations, and meet customers where they are: cautious and alert.
Small updates-clear scripts, verified callbacks, and consistent reporting-compound quickly. The result is fewer losses, more trust, and agents who know exactly what to do under pressure.
Further Learning for Support Teams
Your membership also unlocks:
