Anonymized phone-number sets are used to estimate how many contacts an unregistered number may have
Telegram's automatic algorithms can use anonymized sets of phone numbers to calculate the approximate number of potential contacts an unregistered phone number may have on Telegram. When you open the "Invite friends" interface, Telegram shows the resulting statistics next to your contacts so you can see who could benefit most from joining.
Supporting quotations
Our automatic algorithms can also use anonymized sets of phone numbers to calculate the approximate number of potential contacts an unregistered phone number may have on Telegram. When you open the 'Invite friends' interface, we display the resulting statistics next to your contacts to give you an idea of who could benefit most from joining Telegram.
Qualifications & exceptions
"Anonymized" is Telegram's description; the policy does not explain the method.
Not stated here
The text does not say whether the anonymized phone numbers come from contacts that users sync.
Inference & Profiling · Company practice
Automated algorithms may analyze messages in cloud chats to stop spam and phishing
Telegram says it may also use automated algorithms to analyze messages in cloud chats to stop spam and phishing.
Supporting quotations
We may also use automated algorithms to analyze messages in cloud chats to stop spam and phishing.
Qualifications & exceptions
The stated purpose is stopping spam and phishing. Section 5.6 separately says chat contents are not used for ad targeting or other commercial purposes.
Not stated here
The policy does not say what the analysis produces, whether any results are kept, or how this applies to secret chats, which are described as not readable by Telegram.
Inference & Profiling · Company practice
Aggregated usage data builds features like frequent-contact suggestions, which you can turn off
Telegram says it may store some aggregated metadata to create features that work across all your devices, and may use some aggregated data about how you use Telegram to build useful features. For example, it calculates a rating of the people you message most often to show likely contacts at the top of Search, and a similar rating for inline bots so the app can suggest the bots you are most likely to use. You can turn this off and delete the related data by disabling "Suggest Frequent Contacts" in Settings > Privacy & Security > Data Settings.
Supporting quotations
We may also store some aggregated metadata to create Telegram features (see section 5.5 below) that work across all your devices.
We may use some aggregated data about how you use Telegram to build useful features. For example, when you open the Search menu, Telegram displays the people you are more likely to message in a box at the top of the screen. To do this, we calculate a rating that shows which people you message frequently. A similar rating is calculated for inline bots so that the app can suggest the bots you are most likely to use in the attachment menu
To turn this feature off and delete the relevant data, go to Settings > Privacy & Security > Data Settings and disable
Qualifications & exceptions
The example is a per-person rating of whom you message most; the policy does not define "aggregated".
Not stated here
The policy gives only these examples of features built from aggregated data.
Inference & Profiling · Company practice
Ads in large public channels and search can use content language and an approximate country and city
In large public one-to-many channels (1,000 subscribers or more) and in public search results, ads may be shown based on the language of the content, which may be common to several countries. To keep ads relevant, ads can be shown in a particular country and city, approximated from the country and area code of your phone number or from your IP address. Telegram says this data is never passed to advertisers, no other geolocation data is used and no precise location tracking is involved. This lets advertisers restrict ad visibility to certain countries or regions.
Supporting quotations
In large public one-to-many channels (with 1,000 subscribers or more), and public search results, ads may be shown based on the language of the content – however, that language may be common to several different countries. To keep ads relevant, ads can be shown in a particular country and city, approximated based on the country and area code of a user's phone number or their IP address. This data is never passed to advertisers, no other geolocation data is used, and no precise location-tracking is involved for these purposes. This makes it possible for advertisers to restrict ad visibility to certain countries or regions – for instance, creating an ad for Spanish-speaking audiences but opting not to display it within Spain.
Qualifications & exceptions
Described for large public one-to-many channels (1,000 subscribers or more) and public search results.
Not stated here
The text does not say whether the same approximation is used for ads in other placements, such as bots or stories.
Ads can be chosen from the large public channels you follow
Users who follow a sufficient number of large public one-to-many channels (for example in categories such as Finance, Business & Entrepreneurship or Humor & Memes) may see ads from that category in other large public channels or in bots. Users may also see ads selected on the basis of the channels they follow. Telegram says this data is never passed to advertisers.
Supporting quotations
Users following a sufficient number of large public one-to-many channels (e.g., Finance, Business & Entrepreneurship, or Humor & Memes), may see ads from that category in other large public channels or in other bots as well. Users may also see ads selected on the basis of the channels they follow. This data is never passed to advertisers.