← All tech signals
Governance

The Suspected Number

Operators use AI to judge whether a call is a scam. India’s regulator makes those flags an enforcement trigger, and Korea’s ministry points to AI call analysis in phones.

What is it?

Operators now use AI to judge whether a call is a scam while it rings or is under way. In India, that flag helps trigger enforcement against the sender. Korea’s ministry points to on-device AI that analyses call content in real time.

Why now?

India’s telecom regulator is writing AI output into its rules. TRAI’s September amendments make a provider’s AI flag one half of an enforcement trigger, and a February direction asks operators to share flags across networks within two hours, MediaNama reports.

Who’s building it?

Indian operators under TRAI’s rules, Korean carriers and Samsung, and Virgin Media O2 in the UK.

TRAIIndian operatorsSamsungKorean carriersVirgin Media O2
The full read
Why now, who is building it, where the market goes, and what stays unsettled.
Get weekly →
01 · The market
Who’s in it
Builders — operators’ AI systems and phone makers’ on-device call analysis.
Buyers — customers whose operators screen calls and warn them on screen or by alert.
Gatekeepers — regulators like TRAI, whose rules decide which flags start enforcement.
What’s at stake
Whether an AI flag can carry enforcement weight when the systems behind it use different parameters.
Whether legitimate businesses can be wrongly flagged by AI.
How customer data is shared between networks for AI flagging.
What’s unsettled
How operators’ AI systems with different parameters will work together.
Whether wrongly flagged callers can appeal; the direction names no route.
Whether on-device analysis in Korean phones reaches enforcement the way India’s network-side flags do.
02 · How the demand matureda progression, not a timeline
Novelty
Samsung’s Phone app analyses call content and shows a two-stage alert.
Retention
Virgin Media O2 says it warned customers about more than 1 billion calls.
Willingness to pay
Virgin Media O2 offers Hiya’s Call Defence as a free benefit, ISPreview reports.
Category arc
TRAI turns provider AI flags into part of enforcement, beyond customer warnings.
03 · Where the market goesthe forks, not a prediction
→The flag as evidence: AI flags stay one half of an enforcement trigger, paired with complaints, as in India’s amended rules.
→Flags alone start action, with no named route for wrongly flagged callers to challenge them.
→On-device analysis, as in Korea, keeps the call judgement inside the phone, away from the network.
→Operators’ systems with differing parameters produce flags that regulators weigh differently.
Which fork holds turns on how often a flag proves right when it is later checked, and on whether regulators weigh flags from different operators alike.
04 · The Compliance-as-Feature deskadjacent markets, moving now
Proving You Actually Wore It
AI risk scores screen returns.
AI risk scores and image checks screen returns before refunds
CBS News
Human, Verified
A regulator asks banks to test onboarding against deepfakes.
Hong Kong’s HKMA tells banks to review remote onboarding against deepfakes.
Hong Kong Monetary Authority
The Deepfake Dividend
AI-enabled fraud losses drive demand for verification tools.
$893M in AI-fraud losses in 2025.
SecureWorld
05 · The Wire
01
TRAI makes provider AI flags a trigger for enforcement.
Telecom Regulatory Authority of India ↗
02
Operators must pass AI spam flags to rivals in two hours.
MediaNama ↗
03
Korean carriers and Samsung offer real-time AI call analysis.
The Asia Business Daily ↗
04
Virgin Media O2 says it has warned customers about 1 billion calls.
ISPreview ↗
05
Indian AI systems flagged 24.43 billion calls and texts.
TelecomLead ↗
The panel on this signal
They agreed on
–Under TRAI’s amended rules, three or more complaints in ten days plus an AI flag trigger action against a sender.
–Korea’s ministry points to on-device AI analysing call content.
–MediaNama says the February direction names no route for appealing a wrongful flag.
The ML systems engineer disagreed on
–A count of flagged calls measures what a model suspected; without the share later confirmed, 24.43 billion says little about accuracy.
Know more about the method →
✦How this signal was built
01 · Panel
A Delphi panel of 4 AI analysts, each reasoning independently.
→
02 · Debate
They challenge and revise each other’s reads over two rounds.
→
03 · Evidence
Every claim is tied to a cited, linked source on The Wire.
→
04 · Synthesis
A moderator writes it into one read, disagreement and all.