Technology companies, policymakers and digital safety experts are calling for stronger safeguards as artificial intelligence becomes more widely used in everyday life. At the Trust and Safety India (TASI) Festival 2026 in New Delhi, discussions on October 14 highlighted child protection, age verification, AI-generated deepfakes and the responsibility of technology platforms to prevent misuse.
The discussion brought together representatives from OpenAI, Google and Roblox, alongside policy researchers and other industry participants. A central question was how companies can continue developing AI technologies while ensuring that their products do not expose users, particularly children, to avoidable risks.
The two-day festival, held on October 14 and 15, was organised by the Centre for Social Research and TASI. It provided a platform for industry representatives, policymakers and civil society organisations to examine emerging challenges in digital trust and online safety.
AI Development Raises Questions About Safety and Accountability
As AI systems become more capable, evaluating their potential risks before public release is becoming an important part of technology governance.
Participants discussed the limitations of relying exclusively on voluntary commitments from technology companies. While businesses have commercial incentives to introduce new products quickly, inadequate safeguards can undermine public confidence and expose users to harmful outcomes.
Pragya Misra, Head of Strategy and Global Affairs for India at OpenAI, emphasised the importance of incorporating safety into the development process. Her comments reflected a broader challenge facing the industry: balancing the pace of innovation with the need to establish public trust.
Rohit Kumar, Founding Partner at The Quantum Hub, also raised concerns about competitive pressures that may encourage companies to prioritise rapid product development. He argued for clearer regulatory boundaries rather than relying entirely on commercial incentives to protect users.
The discussion highlighted the importance of accountability throughout the development and deployment of AI systems, rather than treating safety as an issue to address only after a product reaches the public.
Google Highlights Testing Before AI Products Are Released
Pre-release testing was another major topic at the festival.
Snigdha Bhardwaj, Director and Global Head of Generative AI, Trust and Safety at Google, discussed the company's approach to evaluating AI systems for potential safety failures.
One method highlighted was red teaming, a process in which testers deliberately probe a system to identify weaknesses, unsafe outputs and possible avenues for misuse. Findings from these assessments can help developers improve safeguards before making a product publicly available.
The discussion also addressed the need for cooperation between technology platforms. Sharing relevant risk indicators could help companies identify harmful material more effectively, although such arrangements must account for privacy, intellectual property and legitimate competitive interests.
Participants also considered whether emerging products, including camera-equipped smart glasses, require additional safety assessments before wider deployment. Devices that combine cameras, AI processing and real-world interactions can introduce privacy and safety questions that may not be fully captured by conventional software testing.
Child Protection Should Not Depend Entirely on Parents
Protecting children online emerged as one of the most prominent themes of the discussion.
Nicky Jackson Colaco, Vice President and Global Head of Public Policy at Roblox, argued that digital platforms should build protective measures into their default settings instead of expecting parents to identify and activate every available safety feature.
Age verification was part of this discussion. Platforms increasingly need ways to assess whether users are old enough to access particular communication tools, content or interactive features.
However, age-assurance systems also raise questions about privacy, accuracy and the amount of personal information users must provide. Effective implementation requires safeguards that protect younger users without collecting unnecessary data or creating avoidable barriers for legitimate users.
The broader message from the panel was that responsibility for children's digital safety must be shared among technology providers, policymakers, parents and other stakeholders.
Deepfakes Create New Challenges for Digital Trust
AI-generated audio, images and videos can make it increasingly difficult for people to determine whether digital content accurately represents a real person or event.
Akash Pugalia, Chief Digital Officer at TP, highlighted concerns about increasingly convincing synthetic media and its potential to mislead audiences.
Such material can complicate the process of verifying information, particularly when recordings circulate online without reliable context. The challenge extends beyond identifying manipulated content: platforms and users also need practical ways to assess provenance, verify claims and respond when misleading material spreads.
Technical detection tools, content authentication methods and cooperation between platforms may help address these risks. Nevertheless, no single measure can eliminate the problem entirely.
For readers, the issue reinforces the importance of checking original sources, seeking independent confirmation and avoiding the assumption that a realistic-looking video or familiar-sounding voice is necessarily authentic.
AI in Education: Supporting Learning Without Replacing Independent Thought
The panel also considered how growing access to AI tools could influence children's learning habits.
Pugalia raised concerns that students who routinely rely on automated answers may miss opportunities to develop reasoning and problem-solving skills. The issue is not simply whether children use AI, but how these tools are incorporated into their learning.
Misra referred to the importance of learning as a use case for AI in India and highlighted the need to ensure that these systems support education rather than merely provide ready-made answers.
AI can assist with explanations, practice questions and access to information. However, students also need opportunities to evaluate evidence, ask questions and develop their own conclusions.
For educators and families, the practical challenge is to encourage responsible use while helping children recognise the difference between assistance and independent understanding.
Why AI Safety Needs an India-Specific Approach
India's digital environment presents particular challenges for safety systems designed for broad international markets. Users differ in their preferred languages, levels of digital literacy, cultural contexts and access to technology.
Participants stressed that safety assessments should reflect these differences. A system that performs adequately in one language or social context may not respond appropriately in another.
At the same time, safeguards must be proportionate. Measures intended to reduce harm should not unnecessarily restrict access to legitimate information or freedom of expression.
This requires ongoing evaluation, local expertise and collaboration among technology companies, civil society groups and policymakers. It also means recognising that online safety is not solely a technical problem: the effectiveness of safeguards depends on how people actually use digital services.
What the TASI 2026 Discussion Means for the Future of AI
The conversations at TASI 2026 underscored a growing challenge for the technology sector: developing more capable AI systems while maintaining user safety, privacy and public confidence.
Pre-release testing, age-appropriate defaults, better detection of manipulated content and responsible information-sharing were among the approaches discussed. Clearer accountability and cooperation across the industry may also help address risks that individual companies cannot resolve independently.
However, the discussion does not suggest that any single technical measure or regulatory intervention will be sufficient. Effective AI governance will require safeguards that evolve alongside the technology and respond to the needs of different communities.
As AI becomes more deeply integrated into education, communication and everyday digital activities, the ability to combine innovation with meaningful protection is likely to remain a central issue for technology companies, regulators and users alike.