The Ministry of Information and Broadcasting (MIB) has released the draft Telecommunications (Television, Radio and Associated Services) Rules, 2026, for public consultation. Comments may be submitted until July 27. The draft seeks to bring television channels, Direct-to-Home (DTH) operators, Headend-in-the-Sky (HITS) platforms, private Frequency Modulation (FM) radio stations, community radio stations, teleports, television news agencies and Internet Protocol Television (IPTV) services under a single authorisation framework created under the Telecommunications Act, 2023.

A single framework for broadcasting services: The draft creates six categories of authorisation:

• Television channel,
• Television channel distribution service (covering DTH and HITS),
• Teleport,
• News agency for television,
• Private radio service,
• Community radio service.

Authorisations would replace the patchwork of licences, permissions and registrations currently issued under the Indian Telegraph Act, 1885 and related broadcasting guidelines. The rules also provide a migration pathway for existing licensees operating under earlier regulatory frameworks. Entities operating television channels, DTH platforms, HITS platforms, and community radio stations can apply to migrate to the new authorisation regime while their existing permissions remain valid.

IPTV receives formal recognition under the new framework: One of the more significant changes is its incorporation into the authorisation framework.

The draft defines IPTV as a television channel distribution service using Internet Protocol over a closed network. Rather than requiring a separate authorisation process, the rules permit an entity to provide IPTV after submitting a declaration that it already holds either:

• An internet services authorisation under telecom rules; or
• A registration as a multi-system operator under the Cable Television Networks (Regulation) Act.

The draft also places IPTV providers under the same programme code and advertising code obligations that apply to other television distribution services. In addition, IPTV providers must retain recordings of programmes and advertisements for 90 days and furnish them to the government upon request.

Television channels become technology-neutral: the rules shift away from a framework primarily focused on satellite television. A television channel may now operate through either:

• Satellite transmission medium; or
• Terrestrial transmission medium.

The draft broadly defines terrestrial transmission to include wireline and wireless infrastructure, the internet, and other non-satellite broadcasting systems. Channels may also be used to switch between satellite and terrestrial transmission modes, subject to government approval and applicable clearances.

Mandatory public-service programming obligations: The draft goes beyond licensing and places explicit public-service obligations on broadcasters. For television, the rules require at least 30 minutes a day between 6 AM and 11 PM on “themes of national importance and social relevance”, including education, agriculture, health, science and technology, women’s welfare, weaker sections, environmental and cultural heritage protection, and national integration. The government may exempt some channels. The draft also requires channels to report their availability on the landing page to the government and rating agencies.

• Private radio service must operate as a free-to-air service, carry at least one hour a day of programmes of national importance and social relevance, and ensure that at least 20% of daily broadcasts are local content.
• Community radio service must run for two hours a day, remain free-to-air, avoid sponsored programmes except those in the public interest, and include the words “community radio” in the channel identity.
• Community radio stations must also set up an Advisory and Content Committee with local community members, and 50% of the committee must be women.

Operational requirements for broadcasters: The draft introduces several operational obligations. Television channels must commence operations within one year of spectrum assignment and remain continuously operational during the validity of their authorisation.

If a television channel remains non-operational for more than 60 consecutive days, it must inform the government and provide reasons. If it remains non-operational for more than 90 consecutive days, its authorisation will be deemed withdrawn unless the interruption resulted from a government order.

Enforcement and monitoring: The draft gives the Centre broad monitoring powers. It can inspect premises, request recordings, and require broadcasters to retain programme and advertising recordings for 90 days. It can also address ownership changes, which must be reported within 30 days and may require prior permission for a change in control or management. Breaches of the rules fall under the adjudication framework under the Telecommunications Act, with an Adjudicating Officer and a Designated Appeals Committee.

Also read: 

• AsiaSat sends legal notices to Indian government, broadcasters,  after IN-SPACe disallows use of Chinese satellites
• How Illegal IPTV Streaming Pirated Rs 700 Crore Worth of JioHotstar Content
• Explained: What’s inside the Texas Attorney General’s lawsuits against Smart TV content tracking

The post Explained: Public-service programming, IPTV and new compliance rules in MIB’s draft broadcasting rules appeared first on MEDIANAMA.

• Original court order [PDF]
• English translation [PDF]

A German court has ruled that Google is directly liable for false claims made in AI Overviews answers. The ruling raises a key unresolved question about accountability in the AI search era. If AI models hallucinate and produce factually incorrect information, should liability be borne by the model provider, the platform that hosts such AI-generated content, or the user who submits the prompts?

What was the lawsuit about? Two Munich-based publishing companies sued Google after AI Overviews wrongly linked them to scams, subscription traps, and shady business practices. The publishers said the false claims were harming their reputation.

The plaintiffs, whose names have been redacted from the court documents, said they had sent a cease-and-desist letter to Google but never received an appropriate response.

What did the court say in its ruling? The court granted a temporary injunction against Google, restraining the company from spreading false claims about the two publishers through AI Overviews.

Here are the key takeaways from the ruling:

1. AI Overviews are Google’s own answers, not search results: The Regional Court of Munich ruled that Google is a direct infringer in this case since AI Overviews answers are its own content, produced by the company’s own AI model. The court argued that AI Overviews work differently from traditional search results.

“Firstly, search results are not only being generated in whatever order, they are not displayed as links or short previews (snippets). Search query results are being summarized in your own words, according to your own structure,” the ruling said.

In one such instance, AI Overviews began its introductory summary about the two publishers with confident claims like “Yes, [company] is known for dubious business practices,” it then continued with its own independent structure, including characteristics of the alleged fraud and recommendations for users. According to the court, this goes beyond mere presentation of links in search results.

Above all, the court found that the AI-powered overview contained statements “that are not even in the search results.” None of the links cited as the primary source in the AI-generated summaries established any ties between the two publishers and the allegedly shady companies. The court therefore held that these were Google’s own statements.

Google built AI Overviews and offers the service to users. Therefore, Google must bear responsibility for what it says “because it alone has influence over the AI’s offering and the algorithms with which the AI operates.”

MediaNama’s queries sent to Google remained unanswered at the time of publication.

2. Search engine liability rules don’t apply to AI Overviews: Typically, tech firms such as Google and Meta are protected under safe harbour provisions, which grant online platforms legal immunity for user-generated content. However, the court held that previous case law shielding search engine operators from liability does not apply to AI Overviews.

The Munich court examined previous rulings from Germany’s Federal Court of Justice (BGH), which protected traditional search engines from excessive liability. Those rulings argued that search engine operators were only indirect infringers because they merely made third-party content discoverable.

However, the court held that this reasoning does not apply to AI Overviews because it generates “independent, new and substantive” statements by evaluating and combining content from multiple third-party websites. This differs from regular search engines, which simply direct users to external websites.

Additionally, the court argued that AI Overviews “is by no means essential” for using the internet. Traditional search results already help users discover content; AI Overviews is merely an add-on.

3. Court rejects Google’s ‘users can verify themselves’ claim: During the hearing, Google argued that users could check the cited links themselves to determine whether the content of third-party websites matched the AI Overview answers. However, the court rejected this argument.

“If AI Overviews is legally treated as completely unreliable, and all of the displayed links need to be checked independently, then its entire function and benefits would be significantly diminished,” the ruling said.

AI Overviews contain “independently understandable content” with no indication that alternative interpretations may exist or that the content may be unreliable. As a result, users have little reason to verify the underlying search results. This is particularly relevant for people who read quickly, the so-called “front-page readers,” the court noted.

4. Regulatory loophole: The court pointed out that if action is taken only after notification by affected parties, and only in cases of obvious violations, victims would have little effective legal recourse against false claims generated by AI models. In such situations, victims cannot sue the third parties cited as sources because they did not make the false statement. Under existing law, they also cannot effectively sue Google.

In light of this, the court held that Google cannot invoke host-provider protections under the EU’s Digital Services Act or rely on the standard notice-and-takedown framework applicable to search engines.

5. AI Overviews and free speech: An AI-generated opinion is “not the expression of an acquired conviction of the persons expressing it, but the result of an algorithm,” the court said.

The AI-generated output in this case is “primarily an expression of Google’s business activities” and “at most a secondary expression of an interest in being able to freely express one’s opinion and beliefs.”

The court held that the plaintiffs’ personality and privacy rights took precedence over Google’s interests because the AI Overviews statements were unfounded. The summaries linked the two publishers to dubious companies that, according to sworn affidavits, had no connection to them whatsoever.

What must Google do now: The court has banned Google from spreading false claims through AI Overviews about the two publishers relating to scams, links to shady companies, subscription traps, phone calls that allegedly never took place, and claims about their availability. Additionally, Google has been ordered to cover 80% of the plaintiffs’ legal costs. However, the risk of repeat violations remains.

Why this matters: The Munich ruling should not be viewed in isolation. A recent study by SparkToro suggested that 68% of Google searches in the United States ended without a click during the first four months of 2026. This has intensified concerns among news publishers that Google is serving its own business interests through AI Overviews while reducing traffic to their websites. Last year, these concerns prompted the European Commission to launch a probe into AI Overviews to examine whether the feature violated laws such as the Digital Markets Act and the EU Copyright Directive.

According to an analysis by AI startup Oumi for The New York Times, Google’s AI Overviews achieved a 91% accuracy rate. While this may appear impressive at first glance, at Google’s scale it could still translate into millions of incorrect answers every hour. Other LLM providers, including OpenAI, Anthropic, Meta, and xAI, could face similar liability if they generate answers based on their own interpretation of source material.

Further, a growing global debate is underway over whether technology companies such as Google should continue to enjoy legal protections under safe harbour laws, particularly as AI systems have increasingly been accused of encouraging self-harm, spreading misinformation, and even contributing to public unrest.

How social media liability works in India: In India, technology companies are generally treated as intermediaries rather than publishers. Under Section 79 of the Information Technology Act, 2000, social media intermediaries enjoy safe harbour protections. Platforms are not liable for user-generated content, provided they act upon obtaining actual knowledge of illegal content and remove it within the prescribed timeframe.

Earlier this year, the Ministry of Electronics and Information Technology (MeitY) also introduced a three-hour content takedown rule under the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026.

The IT Amendment Rules were officially notified on February 10, 2026, and require platforms such as YouTube, Meta, and X to take down flagged content within three hours of receiving a government notice as opposed to the previous 36-hour window. Notably, MeitY did not include the three-hour content takedown provision in the draft IT Amendment Rules released on October 22, 2025.

Also read:

• Google AI Overviews Reduce Click-Through Rates By 58%, Study Finds
• Google’s AI Overview Faces EU Scrutiny Amid Concerns Over Media Fairness
• Google’s AI Overviews Hit 1.5 Billion Users as Monetization Grows: Takeaways from Q1 2025 Earnings Call

The post German court holds Google directly liable for false AI Overviews claims: What it means for AI liability appeared first on MEDIANAMA.

What prompted it: Shinde urged Sarnaik to take strict action against app-based taxi and ride-hailing services, calling the practice of forcing extra payments “nothing less than holding customers to ransom and exploiting them during times of need”. Good service should earn voluntary tips, he said, “not be used as a tool for coercion or blackmail”.

What the complaints allege: Passengers say, drivers:

• Demand additional payments or tips after accepting bookings.
• Threaten to cancel or refuse rides when passengers reject their demands, leaving them stranded.
• Harass and pressure commuters, particularly during emergencies or at odd hours.

What the minister committed to: Sarnaik said the Transport Department would implement measures to curb the “arbitrary functioning” of app-based taxi companies and that the state would soon introduce a new aggregator policy. He thanked Shinde for bringing the matter to the government’s notice.

This is not Maharashtra’s first move on aggregators: The state has steadily built an aggregator regulatory framework.

• May 2025: It notified the Aggregator Cabs Policy 2025, covering Ola, Uber and Rapido, with rules on ride cancellations, surge pricing and driver earnings, including a penalty on drivers who cancel without valid reason.
• October 2025: It published draft Motor Vehicle Aggregator Rules, 2025 under the Motor Vehicles Act, 1988, capping surge fares at 1.5 times the base fare set by the Regional Transport Authority.
• March 2026: Sarnaik revoked the provisional bike-taxi licences of Ola, Uber and Rapido over regulatory violations.

The forced-tipping issue is already under central scrutiny: The practice Sarnaik flagged maps onto the “advance tip” mechanism that the Central Consumer Protection Authority (CCPA) has investigated since 2025.

The CCPA issued a notice to Uber for “forcing or nudging” users to pay advance tips for faster service, which Union Consumer Affairs Minister Pralhad Joshi called an unfair trade practice, and expanded the inquiry to Ola and Rapido. A LocalCircles survey cited by the coverage found that 78% of app-based taxi users still faced intrusive prompts, including advance-tip nudges, despite months of scrutiny.

Why this matters: The action raises a jurisdictional question. Forced tipping through app prompts is a dark pattern, which falls under the CCPA’s remit under the Consumer Protection Act, 2019, not obviously under a state transport department, whose tools cover licensing and vehicle rules rather than app-interface design.

Maharashtra is stepping in to address a problem that the CCPA has investigated since early 2025, yet LocalCircles data show 78% of users still face these prompts, suggesting that central scrutiny has not changed how platforms behave. A second regulator may add pressure, or it may fragment accountability across two authorities that hold overlapping claims and different powers.

The complaints also blame drivers, but the platforms design the advance-tip nudge directly into the app. Whether Maharashtra’s action can reach that design layer, rather than only individual driver conduct, will determine whether it fixes the root cause or only the symptom.

Also read:

• How Advance Tipping Is Quietly Reshaping India’s Ride-Hailing Experience
• Centre Orders Consumer Protection Authority To Probe Ride-Hailing Apps Over Differential Pricing
• Australia to Ban Dynamic Pricing and Hidden Fees: Here’s How India Compares

The post Maharashtra orders action against app-based taxi aggregators over forced tipping and fare demands appeared first on MEDIANAMA.

How P3P works: The user sets up a UPI mandate, a standing instruction that pre-approves payments up to a fixed limit, and the AI agent transacts within those bounds. The protocol combines three parts:

• UPI’s mandate frameworks: Two existing UPI systems, Single Block Multiple Debit (SBMD), which Pine Labs brands as UPI ReservePay, and One Time Mandate (OTM). These systems set aside a fixed amount in the user’s bank account upfront and debit it later when a condition is met.
• Grantex: A separate identity layer that checks whether the AI agent is genuine, enforces the user’s spending limits, and keeps a record of every payment for the user to review.
• Hypertext Transfer Protocol 402 (HTTP 402): An open web standard, a publicly available technical rule any developer can use, that lets software request and make payments in a machine-readable way. This is what allows any AI agent on any platform to use P3P, not just one company’s assistant.

The user approves the mandate once by scanning a code in their UPI app, after which the agent’s requests debit against the reserved funds without fresh approval. The user can cancel the mandate at any time.

Who is live on it:

• Gullak, a digital gold savings platform, is the first to go live. A user sets a rule such as “buy Rs 500 of gold if the price drops below Rs 16,000 per gram”, approves a mandate once, and the AI agent automatically buys the gold when the price meets the condition.
• Vijay Sales, an electronics retail chain, is in a proof of concept, testing P3P to let an AI agent buy a product the moment it hits a user’s target price.

P3P is currently live on UPI ReservePay only. Pine Labs says cards, net banking, wallets and Equated Monthly Instalment (EMI) options are on its roadmap. Its developer documentation separately lists stablecoins, cryptocurrencies pegged to stable assets such as the US dollar, as a future payment rail.

How P3P differs from Razorpay’s agentic payments: Razorpay has offered agentic UPI payments since October 2025, first with the National Payments Corporation of India (NPCI) and OpenAI on ChatGPT, and later with Anthropic on Claude. In those cases, the payment still requires the user’s final consent before completion, even though the shopping and checkout happen inside a specific AI assistant such as ChatGPT or Claude.

P3P uses the same underlying UPI mandate mechanism, but operates as a standalone protocol with a common set of rules that any software can plug into, allowing any AI agent on any app or website to use it instead of tying it to one assistant.

Pine Labs’ prior agentic work: This is not Pine Labs’ first move into agentic payments. MediaNama has covered its earlier talks with NPCI on autonomous UPI payments through its Model Context Protocol (MCP) server, a tool that connects AI systems directly to its payment infrastructure, and its subsidiary Setu’s agentic bill-payments service on Claude and ChatGPT. The company has stayed quiet on privacy questions analysts raised about what data AI providers receive versus what Pine Labs retains, and whether the company stores prompts and transaction details or uses them to train AI models.

Pine Labs positions P3P as a consumer-controlled framework where users predefine spending limits, merchants, triggers and validity conditions, while AI agents only execute transactions within those pre-approved boundaries. However, the system still raises unresolved regulatory, security and liability questions. We outline some of them below:

1. Is Pine Labs using the UPI mandate framework for a purpose it was not built for? A UPI mandate, the standing instruction that pre-approves payments, originally supported recurring, scheduled payments to a known merchant: a monthly subscription, a Systematic Investment Plan (SIP) into a mutual fund, or an Equated Monthly Instalment (EMI) on a loan. It did not originally support one-off, event-triggered purchases that an AI agent independently decides and executes.

With P3P, AI agents can buy gold whenever a price condition is met or grab a flash sale the moment it goes live, stretching the mandate well beyond its original design. Pine Labs has not publicly stated whether NPCI created or approved a separate framework for autonomous AI-triggered purchases under UPI mandates.

2. How does P3P square with RBI’s rule that every payment needs an extra security check? The Reserve Bank of India (RBI), in its Digital Payments E-Mandate Framework, 2026, requires that setting up any mandate be verified using Additional Factor of Authentication (AFA), an extra security step such as a one-time password (OTP) or a UPI Personal Identification Number (PIN). Normally, every UPI payment requires the user to enter their UPI PIN at the moment of paying. P3P removes that step at the point of payment: the user approves the ReservePay mandate once, and the AI agent then debits the reserved funds without requiring a PIN each time.

Under the RBI framework, banks can process recurring transactions up to Rs 15,000 without Additional Factor Authentication (AFA) once a user sets a mandate, but they still require AFA for transactions above Rs 15,000. It is not clear how P3P handles a single agent-initiated payment above that limit, given there is no human present to authenticate it.

3. Who is responsible if an AI agent makes a wrong or unauthorised payment? MediaNama’s coverage of Razorpay’s agentic payments launch found no agreement in the industry on who is liable when an AI agent pays incorrectly. Razorpay said agentic shopping “does not rewrite the rules of commercial liability”, meaning the merchant handles a wrong order and Razorpay covers payment-security failures.

Pine Labs markets P3P as producing a cryptographically verifiable receipt for each transaction, which it says supports dispute resolution, but it has not stated who actually bears liability when an agent pays in error. The question matters most for high-value payments that users cannot reverse, such as locking in a down payment before stock runs out.

4. What data does P3P share with AI providers, and on what terms? When an analyst asked what transaction data goes to AI providers versus what stays with Pine Labs, CEO Amrish Rau declined to answer. P3P now extends from bill payments to autonomous purchase decisions, which means more data about what a user buys and when. Whether the user’s prompts, transaction records and spending behaviour are stored, passed to AI providers such as OpenAI or Anthropic, or used to train their AI models, has not been disclosed.

5. Does the stablecoin roadmap clash with India’s position on crypto? Pine Labs’ developer documentation lists stablecoin as a future payment rail for P3P. India does not recognise cryptocurrencies as legal tender, and the Reserve Bank of India (RBI) has repeatedly raised concerns about the financial stability and monetary risks posed by private cryptocurrencies. How Pine Labs intends to settle agentic payments in stablecoin within India’s regulatory environment, and whether it would receive regulatory clearance to do so, is unclear.

MediaNama has reached out to Pine Labs with questions. This copy will be updated when we receive a response.

Also read:

• Pine Labs in talks with NPCI for ‘autonomous’ UPI payments on Model Context Protocol
• Privacy Concerns Linger As Pine Labs Experiments With Agentic AI-Driven Payments: Q3FY26 Earnings
• Q2FY26 Earnings Call: Pine Labs Moves Beyond PoS, Patents Tap and Go Online Payment Tech

The post Pine Labs launches agentic payments protocol, raising questions on UPI rules, liability and privacy appeared first on MEDIANAMA.

The FIFA World Cup is not available on the platform’s free or basic premium tiers. Several users said the company advertised the Rs 799 plan as supporting access on up to three devices at the time of purchase. However, they claimed that Zee later restricted the plan to a single device, hours before the tournament began. 

Zee restores original device limit, issues persist: The company restored the original device limit for its World Cup subscription plans after backlash. However, subscribers continued to report other issues. Several users said Zee had advertised World Cup streaming in 4 K resolution. However, the company later informed them that sports content was not available in 4K, with mobile streaming capped at 720p and desktop viewing limited to SD quality. Meanwhile, fans also complained about persistent buffering during matches and said the platform had not activated some paid subscriptions even after the FIFA World Cup had begun.

Rights acquired shortly before the tournament: FIFA struggled to secure a broadcaster for the 2026 World Cup in India, with the rights remaining unsold until June 1, just days before the tournament began. The governing body initially sought around $100 million for the combined India rights to the 2026 and 2030 World Cups before reportedly lowering its expectations to about $60 million. However, broadcasters remained cautious. JioStar reportedly offered around $20 million, while Sony held discussions but did not submit a bid. Moreover, late-night kick-off times in India reduced commercial appeal. Against this backdrop, Zee secured the rights and subsequently obtained an interim anti-piracy injunction from the Delhi High Court before the tournament commenced.

Dark patterns on OTT platforms: A 2025 LocalCircles survey of more than 95,000 OTT users across 353 districts found that many consumers continue to face billing, cancellation and disclosure-related issues on streaming platforms. 

Key findings include:

• 50% of respondents said OTT platforms made it difficult to cancel subscriptions by hiding or complicating cancellation options.
• 24% said platforms continued charging them even after they had cancelled their subscriptions.
• 53% said they were not informed upfront that some content would require additional rental payments despite holding a subscription.
• 47% reported encountering extra charges during checkout that were not clearly disclosed at the outset.
• 77% said platforms used “forced action” tactics, including requiring additional registrations, app downloads, additional information sought, or acceptance of new terms.
• 86% said platforms used interface designs that made it easier to subscribe than to decline offers or cancel services.

Also read:

• Delhi HC Directs ISPs, Domain Registrars to Suspend 79 Rogue Sites Streaming UEFA Champions League
• Over 10-15 OTT Platforms Under MIB Examination, IT Panel Flags Regulatory Gap
• Consumer protection authority fines PhysicsWallah and McAfee for dark patterns: what counts as consent now?

The post Zee draws online backlash for changing device limit on FIFA World Cup subscription plans appeared first on MEDIANAMA.

Coinbase has launched a service that allows AI agents to access user accounts to trade, make payments, and run workflows within user-defined limits. This service is now available through MCP (Model Context Protocol) integration and a CLI (Command-Line Interface).

How does it work? Users can link agents to an isolated portfolio, which restricts access to other holdings, or to their main account. Coinbase plans to introduce configurable rules covering maximum trade size, permitted assets, and spending caps. All agent payments will be subject to the same transaction monitoring and KYT (Know Your Transaction) checks as the rest of the platform.

Spot and derivatives trading for crypto is now available, with future expansion planned for stocks, index funds, prediction markets, and commodities. Key use cases include automated portfolio rebalancing, efficient deployment of idle cash, and scheduling recurring trades based on historical price trends. The system will also integrate with x402, an agentic payments protocol, enabling agents to pay for data and services.

The MCP option supports web-based agent platforms such as ChatGPT and Claude Web, requiring only a single login and no API keys. In contrast, a CLI with a “skill” is designed for terminal-based environments such as Claude Code or Codex, offering lower token overhead and greater customization. A remote MCP that allows users to sign in with Coinbase without API keys is also planned.

Part of a broader push: This expands on Coinbase’s previous AI initiatives. In 2024, the company launched AgentKit, a toolkit that enables developers to integrate autonomous wallets into their apps, allowing bots to manage and transfer crypto without manual input. In December, Coinbase advanced these efforts by adding an AI assistant to its main app, providing users with trading tips and basic financial guidance tailored to their portfolios and market conditions.

Lincoln Murr, Coinbase’s Head of AI Product, identifies this as the key distinction from competitors such as Robinhood. According to Murr, while most platforms allow agents to trade, Coinbase enables both trading and payments, positioning the company as infrastructure for the “agentic economy” rather than simply another brokerage with an added bot.

Why this matters? Coinbase’s launch comes as financial and payments companies compete to build infrastructure for autonomous agents. Last month, Visa invested in Replit to support agentic payments for developers, and this week formed a separate partnership with OpenAI to explore similar products. These moves indicate that traditional card networks view agent-initiated transactions as a significant new revenue stream, not just a niche experiment.

Mastercard and other payment processors are also developing agentic-commerce initiatives, suggesting that the payments industry is restructuring itself to accommodate AI systems that can initiate purchases and trades at scale.

Also read:

• Coinbase Secures FIU-IND Approval Amid India’s Evolving Crypto Policies
• CCI Approves Minority Shareholding Acquisition By Coinbase In CoinDCX
• MP Raghav Chadha Calls for Cryptocurrency Legalisation and Regulation in India

The post Coinbase launches tools that allow AI agents to trade and spend on users’ behalf appeared first on MEDIANAMA.

On June 10, the European Union released the Code of Practice (CoP) under Article 50 of the EU AI Act, the provision requiring transparency in the labelling of AI-generated content. It has two sections aimed at two different stakeholders:

• Providers: Those who build AI systems; and
• Deployers: Those who use these systems to create and publish content.

For AI Deployers: Obligations for Labelling Deepfakes, Including Text: Under the new guidelines, AI deployers/platforms are obligated to label AI-generated content in the manner specified by the EU. The broad categories of AI-generated content covered under this section are:

1. Deepfakes: AI-generated or manipulated images, audio, or video that “resemble existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful.”
2. Published text on matters of public interest: AI-generated text published to inform the public, “without human review or editorial control and where no natural or legal person holds editorial responsibility for the publication.”

Labeling deepfakes clearly, visibly, at the first exposure: The legal requirement is that AI-label disclosures must appear “at the latest at the time of the first interaction or exposure.” Late disclosure does not comply.

Three EU icons are available and free to use, without attribution. They were empirically tested across Member States for noticeability, recognisability, and clarity:

• “AI GENERATED” — for fully AI-generated deepfakes or published text.
• “AI MODIFIED” — for partially AI-manipulated content.
• Basic “AI” icon — a minimal version that can be supplemented with a custom interactive layer.

The Code notes that user testing found that “the variants that include a clear textual label (i.e. ‘modified’) performed significantly better in terms of noticeability and clarity.”

• Equivalent custom labels are permitted but must follow the same design specifications. The main visual element must be the capitalised acronym “AI” in English (or a national language if required by law). Proportions must be preserved if resized, and the label must remain clear and distinguishable at any context-appropriate size.
• Label placement for visual content: The icon must appear “in an appropriate place where no intervening overlay elements exist (e.g., in the top right corner of an image or video deepfake).”
  • For video: The label must be displayed at the beginning and at regular intervals, at a minimum after interruptions such as ad breaks.
  • For published text: The label should appear “above or at the top of the text, near the headline… or in the colophon at the beginning of the text.”
• Label placement for audio-only content: A spoken disclaimer must appear at the beginning, “in plain and simple natural language… disclosing the artificial origin of the audio deepfake in a perceivable manner.”
  • For long-form or live audio, reminders must be repeated “at regular intervals… for the entire duration of the audio deepfake.”
  • When a screen is also available (e.g., a car display or smartphone), a visual icon is required in addition to the audio disclaimer.
• Accessibility Requirements: Compliance with the European Accessibility Act and the Web Accessibility Directive is mandatory. The Code requires:
  • Audio descriptions for visual labels;
  • Tactile or haptic cues for audio content (e.g., “a vibration alert before audio playing”);
  • High-contrast icons and screen-reader compatibility; and
  • Detectability by assistive technologies.

What Are the Obligations for Low-Risk Fictional Content?

• Artistic and Satirical Works Are Not Exempt: they are simply given more flexible placement requirements. Deployers must still label such content, but “in a way that does not hamper the display or enjoyment of the work.” The icon must continue to follow the designated design specifications.
• Digital and Interactive Contexts: For websites, apps, smart glasses, and similar interfaces, the icon may be placed “outside but adjacent to the video or image frame, or adjacent to the audio content and integrated into user interface elements or overlays.” A non-obtrusive icon that reveals more information on click or hover is acceptable, provided it is “perceivable by the end-user without the need to perform dedicated actions.”
• Non-Digital Contexts: For galleries, cinemas, festivals, and similar venues, AI disclosures may appear “at the online or physical point of entry or sale, as part of the introductory or accompanying information (e.g., exhibition leaflet or entrance ticket), or information provided via a physical carrier (e.g., packaging).”

 Additional Recommendations for AI Deployers: 

• Documentation: Deployers must document how they implement labelling, proportionate to their size. Documentation may include “a general description and representative, concrete and real examples of how disclosures are implemented in practice.” Publishing this documentation publicly is encouraged.
• Staff Training: Staff awareness training is required and should be proportionate to organisational size. Training should cover: When disclosure is legally required; How to implement disclosures in workflows; Edge cases such as artistic works and editorial responsibility; Accessibility requirements; and Procedures for correcting incorrect or missing labels.
• Fact-Checking and Flagging Channels: The Code encourages mechanisms that allow “trusted flaggers, independent researchers, academics, fact-checkers” and others to report missing or incorrect disclosures. Reported cases of mislabelling must be reviewed and remedied “without undue delay.”

Maintaining Humans in the Loop:

• Licensed media organisations may rely on their editorial processes to satisfy disclosure obligations for published text, provided those processes meet the Code’s requirements.
• Mandatory Editorial Controls: “All other deployers publishing AI-generated text on matters of public interest must establish internal editorial control policies, which must include at minimum:
  • The identification of the natural or legal person with editorial responsibility (name, role and contact details); 
  • An overview of the concrete organisational measures as well as human resources, allocated to ensure adequate human review or editorial control is performed and editorial responsibility is assumed before publication of the published text.”
• Public Accountability: Contact details of the person or entity with editorial responsibility must be publicly available. Deployers must “publish the contact details of the function, the natural persons or the legal persons with editorial responsibility to ensure accountability.”
• Protection of Press Freedom: “The implementation of this Commitment shall in no way affect media freedom, editorial independence and protection of journalistic source information.”

For AI Providers: Marking and Detection

The Core Problem: The Code argues that “AI systems can generate and manipulate large quantities of synthetic content and it is becoming increasingly difficult for humans to distinguish this content from human-authored content,” creating risks of misinformation, large-scale manipulation, fraud, impersonation, and consumer deception.

Two Mandatory Layers of Marking: “No single marking technique suffices to meet the four requirements in Article 50(2) AI Act, namely effectiveness, interoperability, robustness, and reliability… only an appropriate combination of marking techniques and associated detection mechanisms can allow satisfaction of those requirements in a holistic manner.”

The two mandatory layers are:

• Digitally signed metadata: Where content formats support metadata (audio, image, video, PDFs, Word documents, etc.), providers must record whether content is AI-generated or manipulated and digitally sign and timestamp this information “in a secure and tamper-evident manner.” Providers are encouraged to include additional provenance information, such as: AI provider name; AI system name; Timestamp; Model ID; and Model version.
•  Imperceptible watermarking: This is mandatory for all content except “very short text,” currently defined as fewer than 200 tokens. “For free-form text longer than 200 tokens, watermarking still needs to be applied, even though it may have lower reliability compared to that of watermarking very long text.”
  The Code identifies two approaches:
  • Post-hoc watermarking: applied after generation and 
  • Model watermarking: embedded during the inference-level.

Inference refers to the “‘doing’ part of artificial intelligence. It’s the moment a trained model stops learning and starts working, turning its knowledge into real-world results.” The EU encourages model-level watermarking to simplify compliance for downstream developers.

Optional: Fingerprinting and Logging: AI providers may optionally implement fingerprinting or logging mechanisms. However, “relying on fingerprinting or logging alone is not considered sufficient to meet the quality requirements.” The Code clarifies that this measure applies only to how a provider designs and implements its AI system and does not require logging, monitoring, or retaining prompts or user interactions generally.

Preserving Existing Markings: AI providers must “retain, and abstain from intentionally altering or removing, existing metadata markings” when content is reprocessed. They must also prohibit the removal of markings through their terms of use and acceptable-use policies.

Key exception: An exception exists for “good faith, legitimate processing where the modification… is necessary to maintain accurate and functional information following downstream processing or… security audits and research purposes.”

AI Providers cannot sell circumvention tools. Providers must not “place or make available on the market, nor promote or advertise the use of tools whose purpose is to circumvent the machine-readable markings.”

AI Providers are encouraged to embed richer provenance metadata (Optional): Providers are encouraged to include: AI system name; Provider company name; Generation timestamp; Model ID and version; and For manipulated content, the type of modification performed (e.g., object removal). Multiple operations may be combined into a single marker to reduce complexity.

AI Providers are encouraged to build in a labelling tool for AI deployers (optional): AI providers can offer an optional labelling tool for deployers (especially for deepfake or text systems), allowing them to add required labels at their discretion. This is to support deployers’ compliance without shifting responsibility. However, responsibility for labelling remains with deployers.

How the EU plans to make deepfake detection accessible? 

• Free Detection as the Default: “Signatories will make the detection solution available free of charge.” However, providers with fewer than one million monthly users may charge a fee if the detection solution incurs substantial operational costs and a single user exceeds a reasonable request threshold. Any fee must be “reasonable, fair and proportionate.”
• Free deepfake detection access without any limits on volume: “All Signatories will always provide free access to their detection solution, without any restriction on the volume of requests, to competent market surveillance authorities and other regulators, law enforcement authorities, media, fact-checkers, trusted flaggers, independent researchers, educational and research institutions, and civil society organisations.”
• Text watermark detection may be gated to expert users: Because free-form text watermarking remains less reliable, access to some text watermark detection systems may be restricted to verified expert users until more robust solutions emerge.
• Transparency about detection methods: “Signatories will ensure that detection results indicate whether they are based on a metadata marking, a watermark marking, forensic detection or other techniques, to the extent technically feasible.”
• Downloadable Signed Results: Providers must ensure detection results can be downloaded in a “digitally signed format, including at least a hash of the content submitted for detection, a URL or other identifier of the detection solution, and a timestamp.”
• Zero Data Retention: “The content is stored only for the duration of the detection and is permanently deleted immediately thereafter (i.e., with a ‘zero retention’ policy).” Content submitted for detection cannot be used for any other purpose, including AI training.
• Forensic Detection Remains Optional: Detecting AI-generated content without prior markings remains optional and is currently considered immature. “At the time of publication of the Code, forensic detection mechanisms were not deemed mature enough to comply with the quality requirements.” Providers using such systems may restrict results to verified expert users.

The four-tier testing framework: All marking and detection solutions must satisfy four quality requirements:

• Effectiveness: Normal users should be able to understand the results.
• Reliability: AI-content labelling should maintain low error rates across diverse content types.
• Robustness: Labels should survive compression, cropping, paraphrasing, screen recording, and deliberate attempts at removal.
• Interoperability: Solutions should function across different AI systems and platforms.

By February 2, 2027, AI providers should implement at least one of four prescribed interoperability approaches for watermark detection:

• A publicly available interoperable API standard;
• A publicly readable signpost embedded in content;
• A shared AI industry consortium detection solution; or
• Another comparable interoperability solution.

Also read:

• Experts Warn No AI Can Reliably Detect Deepfakes 
• Explainer: Why detecting deepfakes is a challenging problem 
• Event Report- Regulating For Deepfakes In India, Nov 5

The post Lowdown: The EU’s new rules for labelling AI-generated content appeared first on MEDIANAMA.

Snapchat users aged 13-15 will no longer be able to share Spotlight posts with the general public. They can now share only with people they mutually follow. This update also provides a separate profile for this age group, where Stories and Spotlight content are visible only to mutual friends and do not display engagement metrics such as favourite counts.

What Changes, and for Whom?

Previously, Snapchat allowed users in this age group to post Spotlights publicly without linking posts to their profiles, which prevented direct contact from strangers. This option is no longer available for users under 16.

Users must confirm they are at least 13 years old to register. If Snapchat detects an account belonging to someone under 13, the account is deleted. Additionally, users who register with an age between 13 and 17 cannot later change their birth year to 18 or older, which prevents this common workaround.

Snapchat blocks strangers from sending friend requests or messages to teenage users and displays a warning when teens initiate chats with unknown individuals. The platform also limits the types of content available to teenagers.

Legal pressure in the background

The announcement follows increased legal scrutiny of Snap’s impact on young users. Earlier this year, Snap settled a US lawsuit alleging it facilitated social media addiction and is contesting similar cases nationwide.

CEO Evan Spiegel told CNBC that studies show Snapchat has a positive effect by helping users stay connected to existing friends. He argued the platform should not be compared to TikTok and Instagram. However, the timing of the policy changes aligns with ongoing litigation.

Snapchat is not alone in this approach. In recent years, Instagram has also strengthened restrictions like parental approval for under-16s to change any of the built-in protections on teen accounts, which provides parents more control. Regulatory and legal pressure regarding teen safety has made these measures increasingly common across the industry.

These changes coincide with increased global scrutiny of Snapchat’s youth safety measures. Earlier this year, the company publicly criticised Australia’s social media minimum age law, which bans under-16s from major platforms, but agreed to comply. Snapchat argued that age-verification technologies remain inaccurate and can be bypassed and instead recommended age checks at the app store level through Apple and Google. The company reported locking or disabling more than 415,000 accounts in Australia to support compliance. Snapchat also stated it is primarily a messaging service, not a traditional social media platform, and questioned whether it should be subject to these restrictions.

Read more:

• Youtubers Sue Snap For Allegedly Violating US Copyright Law To Train Its AI Models
• Snapchat To End Unlimited Free Photos and Videos Storage: What It Means
• The problem with the demand for app store age verification of children

The post Snapchat locks public Spotlight sharing for under-16 users appeared first on MEDIANAMA.

The Delhi High Court recently directed US-based website-hosting platforms Vercel, GitHub and Netlify to take down URLs infringing on the “IndiaMART” trademark. Court documents seen by MediaNama reveal the modus operandi of fraudsters allegedly operating fake websites, WhatsApp accounts and bank accounts using IndiaMART’s copyrighted works, including its website and app content. But what makes the ruling more significant is that it demonstrates how Indian courts are increasingly holding cloud infrastructure providers like Vercel and GitHub directly accountable for infringing content hosted by third parties, with extremely short compliance windows.

1. A clone and phish operation: B2B marketplace IndiaMART — with 219 million registered buyers, 8.6 million suppliers, and 124 million product listings as of March 2026 — has a verified seller base that fraudsters identified as a high-value target.

The perpetrators created 15 “near-identical” clones of the IndiaMART website, mimicking its layout, trade dress, graphic user interface, search structure, and features down to “call now” and “get better price” action buttons.

These fake sites were weaponized through WhatsApp. Here’s how it works.

• Fraudsters send WhatsApp messages or emails containing links to the fake URLs impersonating IndiaMART’s website to registered sellers.
• When the seller clicks the link, they are directed to a cloned webpage asking them to “anchor registered phone number.”
• Believing the site to be real, the seller enters the number. In the background, perpetrators use this number on the real IndiaMART site to trigger a legitimate login OTP.
• The seller receives the OTP and enters it on the fake site.
• The fraudsters capture the OTP in real-time, gaining unauthorised access to the seller’s account and financial data.

2. Why Vercel and GitHub are in the dock: According to the suit, these fake IndiaMART websites were hosted on app and website hosting platforms Vercel, GitHub, and Netlify.

While Vercel provides cloud infrastructure for building and scaling websites, GitHub allows developers to create, store, and share code and content. As per the court order, both these platforms enabled continued dissemination and accessibility to infringing IndiaMART URLs by permitting their infrastructure and services to be used for perpetrating fraud.

The court ordered:

• Vercel and Netlify to suspend or disable seven and eight specific URLs, respectively, within 36 hours.
• GitHub to block a repository used to store the source code for the fake sites.
• WhatsApp to block five specific phone numbers (ranging from Indian +91 numbers to those with UK, Australian, and Malaysian country codes) and disclose subscriber information.
• Punjab National Bank and Bharti Airtel to disclose the KYC details of the bank account and mobile number registrants involved in the fraud.

3. Why it matters: The ruling signals a shift in how Indian courts view the “safe harbor” protections for cloud infrastructure providers. While intermediaries like internet service providers often claim exemption from liability for user-generated content, the Delhi High Court has drawn a distinction when the infrastructure is knowingly or actively used to commit fraud.

And in doing so, the court treated Vercel, GitHub and Netlify as extensions of the fraudster’s operations, compelling them to act as enforcers.

Further, the order also mandates that the Department of Telecommunications (DoT) and Ministry of Electronics & IT (MeitY) issue notifications to all TSPs and ISPs to block access to the rogue sites, creating a centralised blocking mechanism.

4. Vercel’s legal troubles in the US: This is not the only instance of Vercel landing in a legal soup. Earlier this week, Vercel admitted wrongdoing and agreed to pay the US government a fine after it failed to comply with a federal search warrant issued under the Electronic Communications Privacy Act. The case dates back to August 2025, when U.S. Magistrate Judge Ryan Carson issued a search warrant requiring Vercel to disclose the contents of a specified user account in its possession, custody, and control.

Three days after Vercel received the search warrant, but prior to the company taking any action to execute the warrant, the user deleted the account. Although the deleted information was still located in Vercel’s servers in a deletion queue, it did not recover that information and only provided the government with some records associated with the account. Vercel failed to meet its obligations to produce the entire contents of the account, believing and representing to the court that the records had been deleted.

Also read:

• Calcutta HC refuses to order OpenAI to show IndiaMart links on ChatGPT: what this means
• Explained: IndiaMART vs OpenAI Lawsuit – Who Controls Visibility On The AI-Driven Web
• Q3FY26 Earnings Call: IndiaMART Stops Sharing Traffic Data Amid Rise of Agentic AI, Chatbots

The post Delhi HC orders Vercel, GitHub to remove fake IndiaMART websites, blocks WhatsApp accounts appeared first on MEDIANAMA.

Canada tabled Bill C-34 at first reading on June 10, 2026, proposing two laws in one omnibus bill: the Digital Safety Act (DSA) and the Digital Safety Commission of Canada Act (DSCC Act).

The bill regulates social media platforms, AI chatbots, and other online services in Canada and establishes a new regulator, the DSCC, to enforce the regulations. Penalties reach the greater of about US$7.2 million (C$10 million) or 3% of global revenue. The bill sits at first reading and has not passed.

What the bill covers: Bill C-34 regulates three service types, each above a user count that regulators will set later:

• Regulated social media services carry the heaviest obligations.
• Regulated chatbot services are artificial intelligence (AI) systems that use a natural language interface, can simulate sustained human-like relationships, and generate non-predetermined responses.
• Regulated online services are other sites or apps that the Governor in Council flags as posing a significant risk to children.

What it excludes:

• Telecoms that provide only basic internet connectivity.
• Private messaging features on any platform.
• Search engines, navigation tools, and e-commerce listings.
• Proactive content scanning, which platforms need not do, though regulators may require technology that blocks child sexual abuse material uploads.

What all operators must do: Every regulated service must build in child-protection design features, verify or estimate users’ ages when it serves pornographic content, and maintain compliance records.

What social media platforms must do:

• Cut the risk that users encounter any of the seven harmful content categories, which are described later
• Block under-16s from holding accounts on services regulators specify, unless the DSCC grants an exemption for adequate child-protection safeguards
• Give users tools to block others and flag harmful content
• Label synthetic content, and label harmful content that bots amplify
• Provide a resource person for user concerns
• Preserve violent or terrorist content for one year after it is removed

What chatbot services must do: Chatbot operators must cut the risk that the bot sends harmful content, interrupt the chat and steer users to human crisis help when a user voices suicidal thoughts or an intent to self-harm, and curb four named behaviours: posing as a human, posing as a licensed professional, using manipulative techniques that build emotional dependency, and encouraging self-harm or suicide.

The seven harmful content categories:

• Intimate content shared without consent,
• Content that sexually victimises a child or revictimises a survivor,
• Content that pushes a child toward self-harm, including disordered eating and suicide content,
• Content that bullies a child,
• Content that foments hatred,
• Content that incites violence,
• Terrorism or violent extremism content.

Two carve-outs apply. The hate definition spares content that merely offends, discredits, or humiliates. The child sexual abuse and terrorism definitions spare material that serves a legitimate purpose, such as journalism, art, education, medicine, science, or justice, where it poses no undue risk to children.

What a digital safety plan must contain: Platforms must file and publish a digital safety plan with the DSCC. The plan must cover:

• Risk assessments and mitigation measures, with effectiveness indicators,
• The volume and type of harmful content the platform moderated,
• User flags the platform received and acted on,
• Synthetic content labelling,
• Compliance resourcing, including automated decision-making,
• Any law enforcement notifications.

Operators may strip trade secrets, confidential commercial information, and the underlying data inventory from the public version.

What the DSCC is and how it works: The DSCC is a new federal body of three to five full-time members, appointed for renewable terms of up to five years, and led by a Chairperson who holds deputy-head status. Its mandate covers:

• Enforcing the DSA,
• Investigating complaints about child sexual exploitation content and non-consensual intimate imagery,
• Issuing guidelines and regulations,
• Accrediting researchers to access the platform data.

The DSCC must consult the Privacy Commissioner before issuing age-verification guidelines and coordinate with the Canadian Radio-television and Telecommunications Commission (CRTC) and the Royal Canadian Mounted Police (RCMP).

How complaints work: Any person in Canada can flag harmful content or a chatbot’s harmful behaviour to the DSCC. For child sexual abuse material and non-consensual intimate imagery, a person can file a formal complaint, but only after they exhaust the platform’s own process first. If the DSCC does not dismiss the complaint, it must order the platform to hide the content while it investigates, then order permanent removal if it finds reasonable grounds.

How researchers get platform data: The DSCC can accredit researchers and educators whose work serves the act’s purposes. Once accredited, they can read the data inventories in the platforms’ safety plans, and the DSCC can order a platform to hand over the underlying data itself, under confidentiality, security, and privacy conditions. If a platform violates that order, the researcher can file a complaint with the DSCC. The Commission can publish the names of accredited researchers and the projects it has cleared.

How the DSCC enforces compliance: The Commission holds wide investigative powers. It can summon witnesses, compel evidence, and name inspectors who, under a warrant, can enter premises (including remotely), copy data, and demand assistance. It can order an operator to take or stop an action and register that order with the Federal Court.

The penalty process proceeds through notices of violation, and an operator can instead make representations, pay the penalty, or sign a binding undertaking. The Commission can name violators publicly and publish their undertakings.

Protections built into the bill: Two confidentiality protections stand out. An employee who files a submission can ask the DSCC to shield their identity, and unauthorised disclosure carries criminal penalties. Separately, when the DSCC receives a user’s chatbot inputs and the bot’s replies, it treats them as confidential if the operator and user agreed to keep them private, a notable safeguard given how much of the bill targets chatbots. Operators can also flag trade secrets and confidential commercial information, which the Commission must protect.

What the penalties are:

• Administrative: the greater of about US$7.2 million (C$10 million) or 3% of global revenue.
• Criminal, on indictment: the greater of about US$14.4 million (C$20 million) or 5% of global revenue.
• Criminal, on summary conviction: the greater of about US$10.8 million (C$15 million) or 4% of global revenue.

The bill bars jail time for unpaid fines and says penalties aim to drive compliance, not to punish.

What the bill leaves undecided: Bill C-34 sets up the structure but leaves key specifics to regulations that cabinet and the DSCC will write later, so even after it passes, some basic questions stay open:

• Which platforms are covered? The rules apply only above a user-count threshold, but the bill never sets that number.
• What “child-protection design features” must platforms build in? The bill requires them, but does not describe them.
• How must platforms verify age? The bill mandates age checks but leaves the acceptable methods to regulation.
• Which online services beyond social media and chatbots must comply? This third bucket only fills once the cabinet designates categories.

How this fits the global age-verification wave: Canada’s under-16 account rule joins a fast-moving global push to keep younger children off social media. Australia implemented the world’s first under-16 ban in December 2025, and VPN downloads surged as the codes took effect. France voted for an under-15 ban, while Spain and Indonesia moved on similar measures.

Canada’s bill leans toward a duties-and-design model over a blanket cutoff, placing it among a smaller set of design-first approaches. University of Ottawa law professor Michael Geist called the volume of decisions left to cabinet and the future commission “astonishing,” predicting the law will take years to implement.

Why this matters for India: India is debating its own version of this question. Several states moved first:

• Karnataka proposed an under-16 ban.
• Andhra Pradesh proposed an under-13 ban.
• Goa is examining similar measures.

Legal experts question whether states hold jurisdiction over internet policy. The central government is reportedly preparing a graded national framework with three age brackets (8-12, 12-16, 16-18) rather than a single ban. The Internet Freedom Foundation (IFF) calls such restrictions disproportionate, warning they ignore engagement-maximising platform design and risk deepening India’s digital gender divide.

Canada’s chatbot duties, its researcher data-access regime, and its privacy-aligned age verification all raise questions India’s framework has yet to resolve.

Also read:

• Karnataka CM Proposes Social Media Ban for Under-16s: What Challenges Lie Ahead?
• Why banning children from Social Media avoids the real problem
• How The Data Protection Bill Restricts Children’s Access To The Internet

The post Explainer: What Canada’s Safe Social Media Act means for platforms, chatbots, and children appeared first on MEDIANAMA.