Generative AI is now part of children’s toys. These novel AI-powered toy “companions” are available on popular e-commerce platforms. Their makers claim the toys help educate children. But experts warn that such toys could impact a child’s healthy development.

How do AI toys work?

AI toys require internet connectivity to work and can take on the form-factor of plush aliens, fluffy animals, or friendly-faced robots. Some robotic AI toys can move independently, while those in the form of stuffed animals are meant to be carried around by their owners.

Many of these AI toys come with embedded microphones that listen to children, to formulate replies. Their makers promote them as products that offer educational answers, give emotional support, guide children through tasks or games, teach them new skills, and return compliments.

Take, for example, Grem by Curio. Shaped like a cute, blue alien creature with large eyes, the website noted that the plush toy was voiced and designed by the singer Grimes. It does not require a subscription.

Meanwhile, many Amazon.com reviews for the ‘Miko 3 AI Robot for Kids’ toy focused on how children were fascinated by the robot, with one buyer noting that it could play hide-and-seek with kids, cheer them up, and serve as an intercom for parents. Others complained about the subscription price and poor battery life.

Multiple AI toy makers on Amazon claim that the models powering these interactions are sourced through reputed providers, such as OpenAI’s ChatGPT, and that there are safety measures in place to ensure they maintain children’s privacy and do not initiate inappropriate conversation topics.

These AI toys are starkly different from toys that make the same sounds over and over again when a button is pressed, or those that can only mimic children’s voices and offer a limited set of replies. AI toys are also expensive.

Why are experts warning parents about AI toys?

AI toys made headlines in late 2025 when the advocacy group US PIRG Education Fund reported that Singapore-based FoloToy’s Kumma bear for children encouraged sexual conversations, and also allegedly told users how to access dangerous objects.

The toy previously used OpenAI’s GPT 4o, but the AI company later suspended the developer and the company moved to another tech provider, according to the group.

Common Sense Media, a tech and media ratings organisation, also outlined the risks of AI toys in a report on January 22. AI toys or “companions” that look cute and soft on the outside but use voice-based chatbots to engage children come with “unacceptable risks,” per the organisation.

Common Sense Media cited risk factors that included unhealthy emotional attachments to AI toys, children’s extremely private data being collected, and cases of chatbots not working reliably. Curio’s Grem and Miko 3 were two of the toys it tested.

The organisation pointed out that children aged five or under cannot properly tell humans apart from AI, so such AI toys could harm how these children develop and recognise key relationships. Even older children between ages six and 12 who might understand that AI isn’t real could still use the toys as a replacement for healthy human connections, per the report.

The organisation listed a slew of concerning factors affecting not just children, but their families as well.

For example, AI toys that use subscriptions could harm children if they become dependent on the toys for emotional comfort while parents are unable to keep up with payments.

Furthermore, about 27% of AI toy outputs were inappropriate for children, as per CMC’s testing, and included content about self-harm, drugs, and unsafe roleplay, apart from going into mature topics and sharing risky advice.

“This happens because the underlying AI models are trained on adult internet content, and child-safety layers are added after the fact. These filters are imperfect, and content designed for adults or teens can leak through when children ask questions in unexpected ways,” noted the organisation.

Furthermore, Common Sense Media stressed the importance of giving children privacy during difficult moments and teaching them to cope with everyday frustrations in healthy ways, instead of distracting them with an always-cheerful toy.

The organisation reported that parental insight tools for these AI toys were largely inadequate, and that children’s private interactions were possibly being shared with third parties for further AI training purposes.

Due to hallucination, the AI toys could also provide incorrect responses and confuse their young users, per the report.

How should parents and caretakers treat AI toys?

Common Sense Media acknowledged that AI toys offered some benefits such as stimulating children’s learning, correcting potential bad behaviour, or telling customised stories to young users. The makers of these AI toys also point out that they offer a screen-free playtime experience for children, while encouraging learning and communication.

However, Common Sense Media recommended that parents engage with their children directly and guide them towards more traditional learning experiences, such as non-AI toys, books, museum visits, playdates, family game nights, imaginative play, and art.

“Human interaction is developmentally essential. No AI toy can replace the benefits of reading together with a parent, playing pretend with siblings, building with friends, or learning from teachers and caregivers,” stated Common Sense Media.

“These messy, complex human interactions are where real development happens. AI toys at best supplement these experiences and at worst replace them—and replacement is the greater risk,” it added.

Published – January 31, 2026 04:45 pm IST

China also plans to create a “Space Cloud” by launching space-based artificial intelligence data centres over the next five years, state media reported on Thursday [File]
| Photo Credit: REUTERS

Apple on Thursday said it has acquired ‍Q.ai, an Israeli startup working on artificial intelligence technology ​for audio [File]
| Photo Credit: AP

The AI model, dubbed “Project Genie”, allows users to simulate a real-world environment through prompts with text or uploaded images [File]
| Photo Credit: REUTERS

Google India registered a flat standalone profit of ₹1,436.9 crore in the financial year ended March 2025 due to lower revenue and higher employee and tax expenses, according to a regulatory document shared by market intelligence firm Tofler.

The company logged a profit after tax of around ₹1,425 crore in the preceding financial year.

When contacted, a Google India spokesperson said that the financial numbers of 2025 were not comparable with those of 2024.

“Profit of Rs 1,425 crore for 2024 includes profit from the IT division. The IT division was demerged into a separate company (Google IT Services), so the 2025 GIPL profit numbers do not reflect IT division profit,” the spokesperson said.

The spokesperson further said that the net revenue for 2024 includes adjustment (addition) of ₹229 crore, which pertains to the revenue of fiscal years 2016-17 to 2022-23 but reflected in 2024 based on the BAPA agreement signed with the Indian government.

A note in the company’s financial report said the company entered into a Bilateral Advance Pricing Agreement (BAPA) with the Central Board of Direct Taxes (CBDT) under Section 92CC of the Income Tax Act, 1961, covering transactions pertaining to the purchase of advertisement space and enterprise products from Google Asia Pacific (GAP) in March 2024.

Under the terms of the BAPA, the company has agreed to the arm’s length price in relation to the purchase of advertisement space and enterprise products from GAP for financial years 2016-17 to 2024-25.

“Pursuant to the BAPA in relation to the aforesaid transactions, the company has recognised the additional income of Rs 2,297 for 2016-17 to 2022-23 under the head revenue from operations (net) in its financial statement for the year ended March 31, 2024, and the additional income of Rs 2,297 has been offered to tax by the company.

The revenue from operations of Google India declined by 3.2% to ₹5,340 crore during the reported fiscal from ₹5,518 crore in FY24.

Total revenue of Google India increased by 3.2% to ₹6,116 crore from ₹5,921 crore a year ago, an account of “other income” of around Rs 776 crore.

According to the analysis done by Tofler, the net margin of Google India also declined to 23.49% from 24.06% a year ago.

“The company’s total expenses for the fiscal year were reported as Rs 4,136 crore,” Tofler said.

The company posted a 7.8% increase in employee benefit expense to about ₹2,146 crore in FY25 from ₹1,989 crore in the year-ago period.

The total tax expense of Google India during the reported fiscal increased by 22.6% to around ₹543 crore in FY25 from over ₹442 crore in FY24.

Published – January 31, 2026 12:42 pm IST

Hoffman is the latest in a string of Silicon Valley figures to criticise the US president [File]
| Photo Credit: REUTERS

Read Comments

Lawyers for xAI and OpenAI did not immediately respond to requests for comment [File]
| Photo Credit: REUTERS

Reuters could not immediately verify the report [File]
| Photo Credit: REUTERS

Hackers and other criminals can easily commandeer computers operating ​open-source large language models outside the guardrails and constraints of the major artificial-intelligence platforms, creating ‌security risks and vulnerabilities, researchers said on Thursday.

Hackers could target the computers running ​the LLMs and direct them to carry out spam operations, phishing content creation or disinformation campaigns, evading platform security protocols, the researchers said.

The research, carried out jointly by cybersecurity companies SentinelOne and Censys over the course of 293 days and shared exclusively with Reuters, offers a new window into the scale of potentially illicit use cases for thousands of open-source LLM deployments. These include hacking, hate speech and harassment, violent or gore content, personal data theft, scams or fraud, and in some cases child sexual ​abuse material, the researchers said.

While thousands of open-source LLM variants exist, a significant portion of the ⁠LLMs on the internet-accessible hosts are variants of Meta’s Llama, Google DeepMind’s Gemma, and others, according to the researchers. While some of the open-source models include guardrails, the researchers identified hundreds of instances where guardrails were explicitly removed.

AI industry conversations about security controls ​are “ignoring this kind of surplus capacity ⁠that is clearly being utilized for all kinds of different stuff, some of it legitimate, some obviously criminal,” said Juan Andres Guerrero-Saade, executive director for intelligence and security research at SentinelOne. Guerrero-Saade likened the situation to an “iceberg” that is not being properly accounted for across the industry and open-source community.

The research analysed publicly accessible deployments of open-source LLMs deployed through Ollama, a ‌tool that allows people and organisations to run their own versions of various large-language models.

The researchers were able to see system prompts, which are the instructions that dictate how ‍the model behaves, in roughly a quarter of the LLMs they observed. Of those, they determined that 7.5% could potentially enable harmful activity.

Roughly 30% of the hosts observed by the researchers are operating out of ‍China, and about 20% in the U.S.

Rachel Adams, the CEO and founder of the Global Center on AI Governance, said in an email that once open models are released, responsibility for what happens next becomes shared across the ecosystem, including the originating labs.

“Labs are not responsible for every downstream misuse (which are hard to anticipate), but they retain an important duty of care to anticipate foreseeable harms, document risks, and provide mitigation tooling and guidance, particularly given uneven global enforcement capacity,” Adams said.

A spokesperson for Meta declined to respond to questions about developers’ responsibilities for addressing concerns around downstream abuse of open-source models and how concerns might ⁠be reported, but noted the company’s Llama Protection tools for Llama developers, and the company’s Meta Llama Responsible Use Guide.

Microsoft AI Red Team Lead Ram Shankar Siva Kumar ​said in an email that Microsoft believes open-source models “play an important role” in a variety of areas, but, “at the ⁠same time, we are clear-eyed that open models, like all transformative technologies, can be misused by adversaries if released without appropriate safeguards.”

Microsoft performs pre-release evaluations, including processes to assess “risks for internet-exposed, self-hosted, and tool-calling scenarios, where misuse can be high,” he said. The company also monitors for emerging threats and misuse patterns. “Ultimately, responsible open innovation requires shared commitment across creators, deployers, researchers, and security teams.”

Ollama did ⁠not respond to a request for comment. Alphabet’s Google and Anthropic did not respond to questions.

Published – January 31, 2026 10:44 am IST

Nvidia CEO Jensen Huang has privately emphasised to industry associates in recent months that the original $100 billion agreement was ‌non-binding and not finalised, the report said [File]
| Photo Credit: REUTERS