Memos

Telegram prohibits collecting data for AI use

tam kien duong — Tue, 03 Feb 2026 00:00:00 +0000

Telegram ▪ Terms of Service, Developer Terms ▪ February 3, 2026

Telegram added an explicit prohibition against data scraping that applies to ‘all users, businesses, and third-party services accessing the platform’. The text refers to the Content Licensing and AI Scraping Terms which explicitly targets data acquisition to ‘train, fine-tune, validate or otherwise engage in the development, enhancement, benchmarking or deployment of artificial intelligence, machine learning models and similar technologies’.

Telegram is a messaging app founded by Pavel Durov and his brother Nikolai. It was marketed as a privacy-focused and secure app. However, this change also serves as a reminder that group chats are not end-to-end encrypted. For example, bot developers can access chat histories and ongoing conversations in order to provide their services. Chatbots can use user interactions to gather data and improve their services in a way that is indirectly related to the user (such as training data and fine-tuning). This may lead to the unintentional disclosure of sensitive information further along the AI pipeline. Despite security concerns and the availability of official alternatives, Telegram was notoriously popular with French officials.

A similar modification has also been detected in the developer terms.

X focuses its automated enforcement on child protection in the UK

tam kien duong — Tue, 27 Jan 2026 00:00:00 +0000

X ▪ Community Guidelines ▪ January 27, 2026

X added clarifications regarding the automated enforcement of child protection regulations. This change has only been detected in the UK jurisdiction, and its language and references are specific to this area.

The introduction to the methodology of “automated content enforcement” mentioned the use of “artificial intelligence models” alongside “heuristics, and machine learning algorithms”. This change does not significantly impact the system and mainly updates the latest language.

A short explanation of how the supervised machine learning model operates has been replaced by a longer explanation in dedicated paragraphs. The example mostly focuses on how automated decisions are used to protect children. It insists on using both internal tools and third-party services. Additionally, it notes that “proprietary technology is steadily reducing the burden on people to report this content”.

Two sections have been added on reporting non-compliance with the UK’s Online Safety Act and on issues concerning “proactive technology”. Mostly, it provides links to relevant forms to file reports.

Seven occurrences of the term “content harmful to children” have been added throughout the document.

The majority of these modifications focus on protecting children from exposure to sensitive content and children’s protection. This must be balanced with the creation of synthetic content that depicts children.

X, formerly known as Twitter, is a social media platform owned by Elon Musk. Grok is a generative artificial intelligence service run by xAI, X’s parent company, that is fully integrated into X. In July 2025, a new version of its image and video generation tool was released, including a “spicy” mode that removes protections regarding adult content. Since December 2025, media and regulators have reported on the persistent use of these tools to produce sexually suggestive images of young women or girls, which are then shared through X.

On January 12, 2026, the UK regulator for communication services, Ofcom, launched an investigation into this service. Other regulators, including the EU Commission, are also investigating the same issue.

TikTok rewrites all US terms following the transfer of its USA operations

tam kien duong — Fri, 23 Jan 2026 00:00:00 +0000

TikTok ▪ Commercial Terms, Data Processor Agreement, Developer Terms, In-App Purchases Policy, Law Enforcement Guidelines, Privacy Policy, Terms of Service ▪ January 23, 2026

On January 23, 2026, TikTok revised multiple policy documents in several jurisdictions. The updates established a new legal entity, “TikTok USDS Joint Venture LLC”, as the new “operator of the Platform in the United States”. Most importantly, the terms that applied globally with a few paragraphs defining local specificities, now have a dedicated US version that applies only to that country.

Documents from other jurisdictions (Ireland, the United Kingdom, Australia and Canada) have been less affected. They mainly reflect changes in the service’s contact points in the US and update links to point to new “global” versions of the terms that now differ from the US version. The Developer Terms, however, contain a new section named “TikTok Developer US Data Sharing Agreement”.

TikTok is the object of scrutiny regarding digital rights, election integrity, misleading political information and direct influence over young audiences. But beyond these points, the United States of America have treated TikTok’s case as a matter of national security, in a long-standing conflict with China that culminated with Oracle buying TikTok’s US operations, leading to these changes.

Starlink allows third-party collaborators to use personal information globally for training AI models and removes mentions of not selling user data

tam kien duong — Fri, 16 Jan 2026 00:00:00 +0000

Starlink ▪ Privacy Policy ▪ January 16, 2026

On January 16, 2026, Starlink introduced several changes regarding the use of personal data.

A previous memo emphasized that this policy shifted from the website to the company’s telecommunication services. The changes described by the memo also mentioned that de-identified data are now ‘from and about’ users.

Personal information is now used “to train our machine learning or artificial intelligence models” for various purposes. They are listed in the same document, which includes “to comply in good faith with applicable laws, legal processes, and lawful government requests” and “to use data analytics to debug, optimize, and improve our products and services, marketing, customer relationships and experiences”.

“Third-party collaborators” have been added along with affiliates and service providers. This opens up the ability to share data horizontally with partners, both inside and outside the parent company. Such third-party collaborators are not defined in the document.

Starlink is a service provided by SpaceX, a company owned by Elon Musk. He is the CEO, CTO, and chairman of the company. He is also the founder and CEO of xAI, a company focused on generative artificial intelligence with the infamous Grok. This software has recently gained attention for enabling the creation of CSAM material. xAI is also the parent company of X, formerly known as Twitter. The website was a prominent and popular microblogging platform. Several governments and companies still use it to host their official communications.

Additionally, users now have to opt out of “the use of your data for artificial intelligence model training by third-party collaborators” if they don’t want their data to be processed.

Finally, the set of changes also included the removal of standard text claiming that the service was “not knowingly selling user information”. This leaves the question of whether they do it now.

Bing applies its intellectual property policies to all types of ads, but adult content remains restricted to search

tam kien duong — Thu, 08 Jan 2026 00:00:00 +0000

Bing ▪ Privacy Policy ▪ January 8, 2026

On January 8, 2026, Bing extended the scope of its intellectual property policies to apply to ‘all ad types’, instead of just ‘Search’. The previous version stated that ‘Adult is not allowed in other ad types’, which include ‘native’, ‘display’, and ‘video’. The updated guidelines outline specific standards for ‘counterfeit policy’ and ‘trademark used in the text of the ad or keywords, as well as imagery contained in the creative’. The modification primarily streamlines the implementation of this policy across the various ad networks that Bing is operating.

However, the adult content policy remains in place for search. The document specifies that adult advertisements are defined by the content of the ad copy and the target website. It also provides a list of countries that may be targeted.

Search ads appear in search results on sites that use Bing, such as DuckDuckGo and Ecosia. This channel is not only limited to text, but also includes images. Search results being limited by the keywords used during interactions with the search engine makes them more reliable in terms of consent. On the contrary, other types of ads are prone to unsolicited displays of content. This partially solves the problem of displaying age-inappropriate content to young audiences. It also ensures that the search engine continues to serve as a conduit for web traffic through paid advertising.

The adult content policy makes it mandatory for concerned ads to be mediated by an age-gating mechanism on the publisher’s responsibility.

Age verification is a controversial issue in many jurisdictions. The efficacy of shielding youth audiences, the practicality of implementing such measures, and the potential for perpetuating surveillance systems are all matters of concern.

Shein specifies that the return window depends on third-party sellers

tam kien duong — Wed, 07 Jan 2026 00:00:00 +0000

Shein ▪ General Conditions of Sale ▪ January 7, 2026

Shein updated its return and refund policies, providing more specific details in its general conditions of sale. It clarified that the return period will vary based on third-party sellers, if applicable. This language is typical and consistent with other similar documents.

Return and refund rules are strictly defined in Europe. In contrast, in the United States, they are more flexible and demand clarity only as a best practice.

Shein is a rapidly growing Chinese online marketplace that initially specialized in fast fashion products but also extended into other kinds of goods. In the context of low-cost fast fashion, return and refund policies are crucial due to lax quality control standards and misleading product pages. Shein also attempted to open a physical store in Paris, which sparked significant controversy.

Telegram expands its age eligibility requirement to Australian users

tam kien duong — Wed, 07 Jan 2026 00:00:00 +0000

Telegram ▪ Terms of Service ▪ January 7, 2026

Telegram expanded its age eligibility requirement by adding Australia to the regions where users must be at least 18 years old to sign up, alongside the EU and the United Kingdom.

The wording was also revised. Whereas the previous policy referred to “citizens of EU countries and the United Kingdom,” linking the restriction to nationality, the updated text applies to “users,” making the requirement dependent on the user’s location rather than citizenship.

Australia’s Online Safety Act 2021 introduced age verification. Telegram was fined AUD 957,780 for non-compliance in February 2025. Since November, Telegram has been challenging the fine on technical, legal and freedom of speech grounds.

Telegram is a messaging application founded by Nikolai and Pavel Durov, two Russian entrepreneurs. Telegram has come under intense scrutiny around the world for being a platform that facilitates many illegal activities, including the sale of illegal pornography and the drug trade. Marketed as a privacy-focused messaging app, it also benefits Durov’s anti-Putin reputation and the company’s “refusal to hand over data in response to lawful orders”. Since then, and especially since Durov’s arrest by the French authorities, Telegram has started cooperating with government agencies and modifying its policies.

YouTube allows viewers in Taiwan to buy Jewels

tam kien duong — Tue, 06 Jan 2026 00:00:00 +0000

YouTube ▪ Content Monetisation Policy ▪ January 6, 2026

On January 6, 2026, YouTube expanded its Virtual Items Program to viewers in Taiwan. Previously, eligibility was described using the term “US viewers”, which was ambiguous and did not clearly specify whether it was based on residency or location. The updated language clarifies that eligibility is based on the user’s location.

The Virtual Items Program is a microtransaction mechanism that allows viewers to purchase bundles of Jewels, which they can transform into digital items. These items can then be sent as tokens of ‘appreciation in real-time’ during vertical live streams. Items received by the creator, as Rubies, are converted into Google AdSense credit at a rate of 1 Ruby per USD0.01. Google AdSense is an online advertising program provided by Google platform.

AliExpress allows the sale of decorative plants in Poland, France, Germany, Italy and Spain

tam kien duong — Sun, 04 Jan 2026 00:00:00 +0000

AliExpress ▪ Marketplace Sellers Conditions ▪ January 4, 2026

On January 4, 2026, AliExpress modified its Marketplace Sellers Conditions to allow local sellers in Poland, Germany, France, Italy, and Spain to sell live ornamental plants. The policy now permits these sellers to ‘sell live plants within the country if they meet the compliance requirements’. An announcement released on October 14, 2025 stated that ‘plant seeds, living plants only for ornamental purposes rather than food purposes’ are authorized. ‘Agricultural and vegetable’ products are still prohibited for sale

AliExpress is a serious competitor for Amazon in the field of online shopping. The Chinese company has recently started delivering fresh food in South Korea and in London.

In November 2023, the European Union was already investigating AliExpress for distributing ‘illegal products such as fake medicines, non-compliant food, and ineffective dietary supplements’.

Facebook now uses AI interactions to personalise experiences and ads

Saumyaa Naidu — Sun, 28 Dec 2025 00:00:00 +0000

Facebook ▪ Privacy Policy ▪ December 28, 2025

Facebook included AI at Meta as a Meta Product to which its privacy policy now applies. The generative AI features and experiences across Meta Company Products are collectively referred to as AI at Meta. The platform will collect information from interactions with AI at Meta, and also share it on Meta Products and with integrated partners. These interactions will include prompts, responses received, actions requested, and feedback given to AI. Facebook will also record information provided by others about a user to AI at Meta, including messages and content that the user may have shared.

In the revised policy, information collected across Products will be used to personalise features, content, and ads in AI at Meta, among other Meta Products. The information will also be shared with other Meta Companies to personalise features and integrations.

Furthermore, interactions with AI at Meta and related metadata such as content and messages will be used for purposes such as personalising and improving Meta Products, promoting safety, integrity, and security, providing measurement, analytics, and business services, personalising ads on Meta Audience Network, undertaking research for social good, and sharing information with others, including law enforcement.

Information collected from user activities on Meta products including posts, comments, audio etc, interactions with AI at Meta, and information from partners, vendors, and third parties, will all be used for the newly introduced purpose of developing and improving AI at Meta. Facebook removed the clarification that it will not share personally identifying information, such as username, unless included in messages to the AIs, with third-parties.

Besides the updates in use of data from AI at Meta, Facebook has also expanded information collection on users’ connections and interactions both on and off Meta Products, and on the online or in person use of third party products and services. The platform added data providers as a third-party public source for datasets which will now be additionally used for undertaking analytics.

The policy no longer provides details of information used to decide which ads and commercial content to show. It also removed factors that influence the personalisation of Facebook Feed and links to manage it. The policy clarifies now that search engines and other third parties use public content about a user (name, username, profile picture, etc.) according to their own terms and policies, not Meta’s. Finally, for non-account holders, Facebook added that it can remove content that violates their terms and policies.

Meta is among the several major platforms that have introduced updates related to AI data use, including Google, LinkedIn, and Open AI. Given this shift, researchers have highlighted concerns such as privacy risks, manipulation using vulnerabilities, and a lack of users’ control over their data and experience on platforms.

TikTok updates its Developer Terms to introduce Commercial Content Library access, clarify research purposes, and reorganize core definitions

Asma Sifaoui — Fri, 26 Dec 2025 00:00:00 +0000

TikTok ▪ Developer Terms ▪ December 26, 2025

TikTok updated its Developer Terms of Service with targeted but consequential revisions that clarified access conditions, introduced a new data category tied to the EU Digital Services Act, and restructured core definitions governing developer use of TikTok services.

The update strengthens the acceptance clause by replacing permissive language with an explicit prohibition. Where the prior version stated that users who do not agree to the terms should not access the Developer Services, the revised text now states that users who do not agree are not permitted to access or use them. This change sharpens the contractual boundaries around consent and access.

TikTok also introduces a new defined dataset, “CCL Data,” referring to advertising and commercial content made available through the Commercial Content Library API. Alongside this, the company adds a new definition of “CCL Purposes,” explicitly framing the use of this data as supporting supervision and research into risks associated with online advertising, including illegal ads, manipulation, and disinformation with foreseeable impacts on public health, security, civic discourse, political participation, and equality. The Commercial Content Library API itself is newly defined and explicitly linked to Article 39 of the EU Digital Services Act, signaling alignment with statutory transparency obligations.

In addition, the Definitions section is reorganized and renumbered without materially changing the substance of most existing terms. Definitions for Developer Content, End Users, TikTok API, Developer Documentation, Developer Services, Sample Code, SDKs, TikTok Services, and TikTok Information are reinserted in a revised order, with minor wording adjustments but no substantive expansion of rights or obligations.

Overall, the changes do not broaden developer permissions but instead formalize a compliance-oriented data access pathway tied to EU law, clarify enforcement language around acceptance of the terms, and structurally reorganize definitions to accommodate the newly introduced Commercial Content Library framework.

Instagram begins use of AI interactions to personalise experiences and ads

Saumyaa Naidu — Tue, 16 Dec 2025 00:00:00 +0000

Instagram ▪ Privacy Policy ▪ December 16, 2025

Instagram included AI at Meta as a Meta Product to which its privacy policy now applies. The platform will collect information from a user’s interactions with AI at Meta, and information provided by others about a user to AI at Meta, including messages and content that the user may have shared. In the revised policy, information collected across Products will be used to personalise features, content, and ads in AI at Meta, among other Meta Products. The information will also be shared with other Meta Companies to personalise features and integrations.

Information collected from user activities on Meta products including posts, comments, audio etc, interactions with AI at Meta, and information from partners, vendors, and third parties, will all be used for the newly introduced purpose of developing and improving AI at Meta. Instagram removed the clarification that it will not share personally identifying information, such as username, unless included in messages to the AIs, with third-parties.

Besides the updates in use of data from AI at Meta, Instagram has also expanded information collection on users’ connections and interactions both on and off Meta Products, and on the online or in person use of third party products and services. The policy no longer provides details of information used to decide which ads and commercial content to show. The policy clarifies that search engines and other third parties use public content about a user (name, username, profile picture, etc.) according to their own terms and policies, not Meta’s.

Pinterest is replacing pharmacies in its list of prohibited ads with casinos

tam kien duong — Sat, 13 Dec 2025 00:00:00 +0000

Pinterest ▪ Commercial Terms ▪ December 13, 2025

In a minor modification to its Commercial Terms, Pinterest replaced an advertising category that was prohibited from being shown in Qatar and Saudi Arabia. ‘Brick-and-mortar casinos’ are replacing ‘brick-and-mortar pharmacies’.

Pinterest permits alcoholic beverage advertisements

tam kien duong — Fri, 12 Dec 2025 00:00:00 +0000

Pinterest ▪ Marketplace Sellers Conditions ▪ December 12, 2025

Pinterest updated its Marketplace Sellers Conditions to permit the promotion of alcoholic beverages in its shopping ads. The change mentions that alcohol advertising must comply with local jurisdictions, such as those regarding legal age restrictions. It also includes a comprehensive list of countries where such advertisements are prohibited. Lastly, it sets down global rules outlining the prohibition of targeting minors and restricting positive alcohol consumption discourse.

Wikipedia clarifies rules for accessing and disclosing temporary account IP addresses

Asma Sifaoui — Thu, 11 Dec 2025 00:00:00 +0000

Wikipedia ▪ Privacy Policy ▪ December 11, 2025

Wikipedia updated its Privacy Policy section on the use of temporary account IP addresses, specifically on refining how access, disclosure, and misuse of this information are governed. The changes do not expand who can access IP addresses, but they clarify permissible uses, add procedural detail, and tighten safeguards around disclosure.

The update reiterates that temporary account IP addresses may only be used for investigation, discovery, and enforcement related to vandalism, abuse, harassment, spam, and other policy violations. It explicitly allows the use of IP auto-reveal functionality during investigations, provided the automatically revealed IP addresses are used solely for these enforcement purposes. The revised text also removes language suggesting that access must be strictly “as-needed,” replacing it with a clearer emphasis on purpose limitation and compliance with applicable local policies.

Wikipedia also refined restrictions on disclosure. While users with access rights may still make limited disclosures when blocking fake accounts, also referred to as “sockpuppets” within the wikipedia community, or performing administrative actions, the policy now explicitly prohibits directly connecting temporary accounts to their associated IP addresses in block reasons or similar public contexts. Disclosures are framed as permissible only when reasonably necessary for enforcement or investigation, and IP addresses must be removed if later determined to be unnecessary or disclosed in error.

The update also clarifies consequences for misuse. An unnecessary or mistaken disclosure alone is no longer framed as grounds for immediate removal of access. Instead, the policy specifies that repeated patterns, intentional misuse, or local community consensus are required to justify revoking access to temporary account IP addresses. Parallel language is added for users with manually granted access, aligning standards across access pathways.

Snapchat adds minimum age requirements in Australia

Saumyaa Naidu — Tue, 09 Dec 2025 00:00:00 +0000

Snapchat ▪ Privacy Policy ▪ December 9, 2025

Snapchat added Social Media Minimum Age (SMMA) requirements of the Online Safety Act 2021 for users in Australia. In order to comply with the SMMA, where “users under 16 years of age must not hold accounts on certain services”, the platform will verify the country of residence and age of the user based on information collected. This would include date of birth, IP address, behavioural information about interaction with Snapchat, information about content created, interactions with My AI (Snapchat’s chatbot), metadata, and information about friends on the platform. Snapchat further requires users to undertake age verification measures using a third-party provider called k-ID. Snapchat will only receive a “yes/no” result on whether the user is above 16, based on which access will be granted. For users below 16, the account will be locked and Snapchat will not receive any personal information provided during the k-ID verification process.

The Social Media Minimum Age (SMMA) law came into effect on December 10, 2025, in Australia and has garnered criticism from teenagers, activists, and politicians. Reddit has also filed a lawsuit against the Australian government to oppose these requirements.

TikTok will not be accessible to children below 16 in Australia

Saumyaa Naidu — Tue, 09 Dec 2025 00:00:00 +0000

TikTok ▪ Privacy Policy ▪ December 9, 2025

TikTok added minimum age requirements in compliance with the Social Media Minimum Age (SMMA) law under the Online Safety Act 2021 in Australia. Under the terms for Australian residents, the platform specified that “TikTok is not directed at children under the age of 16”. The policy also includes links to report users below the age of 16, and to raise complaints and escalate to the Office of the Australian Information Commissioner (OAIC) if needed.

X adds minimum user age of 16 in Australia

Saumyaa Naidu — Tue, 09 Dec 2025 00:00:00 +0000

X ▪ Community Guidelines ▪ December 9, 2025

X specified that users only above the age of 16 can use an X account in Australia. The update has been made in order to comply with the Social Media Minimum Age (SMMA) requirements of the Online Safety Act 2021 under Australian law. The SMMA prohibits certain age-restricted social media platforms from letting people under the age of 16 create or hold an account. While outlining its age-verification approaches, X has listed using existing information such as self-attested age, ID verification, legacy verified public figures, account creation date, and verified government or organisational accounts. When these are unavailable, X will use email address-based estimation to verify the age. If these methods are unsuccessful, the platform will require users to submit a government-issued ID, which will be used to verify their age by artificial intelligence systems.

X has also added terms to handle the information collected during the verification process. It will use ID details for age verification purposes only and destroy the data collected within 31 days. When an account holder is determined to be below 16 years of age, they will be unable to use an X account. Further, the policy details out mechanisms to appeal an age assurance decision and to report a suspected underage account. Similar to the previous version of the policy, accounts belonging to known minors between the ages of 16 and 18 years will be allowed ,with certain restrictions on visibility of posts and receiving direct messages.

YouTube details out criteria for educational, artistic, and public interest exceptions to violating content

Saumyaa Naidu — Thu, 04 Dec 2025 00:00:00 +0000

YouTube ▪ Community Guidelines ▪ December 4, 2025

YouTube expanded on its criteria for exceptions to content violating community guidelines. It clarified that in some cases, content with educational, documentary, scientific, or artistic (EDSA) context including public interest content will be considered an exception to otherwise prohibited content. This was updated in several sections of the community guidelines including Nudity & Sexual Content Policy, Child Safety policy, Suicide, Self-Harm, and Eating Disorders policy, Harmful or Dangerous Content policy, Hate Speech policy, and Misinformation policies.

While the EDSA exception was already stated in many of these sections, YouTube also added public interest content such as “congressional or parliamentary proceedings, campaign speeches, or debates over ongoing government actions” to the list of exceptions and age-restricted content.

Starlink globalizes its policies and adds specificities for Nigeria and South Korea

tam kien duong — Wed, 26 Nov 2025 00:00:00 +0000

Starlink ▪ Privacy Policy ▪ November 26, 2025 - December 4, 2025

On November 26, 2025, Starlink updated its privacy policy to become a global privacy policy. Long overdue, the last official version was from May 1, 2023. The updated policy applies to all services provided by Starlink, not just its website. It now allows using de-identified data ‘from and about’ users to be used for ‘any purpose, including improving our services, to the extent permitted by applicable law’. Data will be disclosed to government regulators or law enforcement agencies in order ‘to protect any person from death or serious bodily injury’.

Additionally, specific privacy policies for Nigeria have also been incorporated. On December 4, another change introduced a dedicated set for South Korea.

These modifications occur in the backdrop of the involvement of US tech companies as an extension of US diplomacy. Starlink is a satellite internet service provider operated by Starlink Services, LLC, a subsidiary of SpaceX, a US company owned by Elon Musk.

Rest of World, an independent media focusing on tech and power in global majority, has previously highlighted the link between Trump’s attention and new Starlink contracts with other countries. Nigeria has been under scrutiny due to allegations of persecution against Christian communities. The United States has threatened to intervene. In late October 2025, President Donald Trump made a visit to South Korea. Since then, South Korean airlines have signed contracts with Starlink to offer in-flight internet to their passengers. Additionally, home internet services were introduced to the South Korean household market at the start of December.

The US Department of Defense began funding the provision of Starlink service to Ukrainian forces during the ongoing Russian invasion of Ukraine started in February 2022. But it also came with some controversies of intervention when Starlink decided to shut down the service during a Ukrainian strike.