India AI Impact Summit 2026

safe-trusted-ai-at-scale

6 min read

TRACK:Safe and Trusted AI
SPEAKERS
  • Arvind Kumar, STPI
  • Ashok Meena, Ministry of Jal Shakti
  • Dr . Piyush Singla, Govt of J&K
  • Dr Rajiv KN, BWSSB
  • Dr. Ravi Gupta, Elets Technomedia
  • Mohammed Safirulla, India AI Mission
  • Sid Sheth, d-MATRIX
SESSION OVERVIEW

As AI adoption accelerates from isolated pilots to enterprise-wide integration, attention is shifting toward the creation of safe, trusted, and scalable ecosystems. This high-impact session examines practical implementation frameworks, governance architectures, and leadership strategies that reconcile rapid innovation with accountability and risk management. The emphasis is on operationalising responsible AI in ways that deliver measurable outcomes while safeguarding security, regulatory compliance, and long-term resilience. By convening policymakers, industry leaders, and technology experts, the session advances a systems-level understanding of how AI can be deployed at scale to drive productivity, institutional trust, and inclusive growth in an increasingly AI-driven global economy. ... Read More

VIDEO RECORDING

Safe & Trusted AI at Scale

Abstract

The panel examined how large‑scale AI deployment can be made safe, trustworthy and compliant with societal values. Drawing on the European AI Act’s “code of practice” model, the participants argued for a flexible yet concrete governance framework that involves governments, industry, academia and civil society. They stressed the need for clear risk‑mitigation rules, coordinated international cooperation, and a focus on context‑specific use‑cases to build both productivity and public confidence. The discussion closed with a call for decisive action from policymakers and the wider AI ecosystem to institutionalise safety standards, especially in high‑risk domains such as military applications and loss‑of‑control scenarios.

Detailed Summary

The moderator introduced the theme of operationalising responsible AI at scale. The opening reference was to the European AI Act’s “code of practice”, described by Professor Benjo as a co‑legislative instrument that brings together civil society, developers (small, medium and large), and academia. The purpose of this code is two‑fold:

  1. Prevent existential and systemic risks (e.g., attacks on democratic processes, misinformation, cyber‑bullying, criminal misuse).
  2. Create a culture of restraint within AI systems, ensuring that innovation does not erode fundamental human rights.

The panel stressed that a clear, non‑vague framework is essential; rules must be specific enough to be enforceable while retaining sufficient flexibility to adapt to rapid technological change.

2. The Role of Government in Enforcing the Code of Practice

2.1. Parliamentary Commitment

A speaker (identified as Brando in the transcript) recapped Professor Benjo’s concluding comment that the parliament will provide the European AI Office with all necessary tools to implement the code of practice. The argument was that although many private firms already comply with many risk‑mitigation clauses, public authorities must match the capacity of powerful private actors to ensure equal footing in rule‑making and enforcement.

2.2. Building Public Trust

The speaker asserted that the ultimate objective is to build trust among citizens:

  • “We can innovate without sacrificing human rights and protection of our fundamental values.”

This trust‑building is portrayed as the foundation for wider adoption of AI at scale.

3. Recommendations for Leaders – A One‑Minute Pitch

The moderator asked Sean (a panelist) to distil a single recommendation for summit leaders. Sean’s response highlighted three inter‑linked points:

  1. Create conditions for safe, beneficial AI development – acknowledging these conditions are currently insufficient.
  2. Enable firms to take additional safety steps despite competitive and geopolitical pressures.
  3. Facilitate global cooperation – not only among European and US firms but also with Chinese developers, ensuring all parties sit at the table as equals.

Sean framed the need for a “red alarm bell”: if companies cannot safely slow down or share expertise, the risk landscape escalates.

4. Focusing on Context‑Specific Use‑Cases (Paula’s View)

Paula shifted the conversation to the gap between perception and deployment of AI’s power. Key points:

  • Trust is context‑dependent – what works in medicine differs dramatically from customer service.
  • Identifying the right use‑cases and trust‑controls within each domain is the lever for unlocking both productivity and trust.

She argued that domain‑specific standards will accelerate adoption while maintaining safety.

5. Balancing Safety with Diffusion (Unnamed Speaker)

Another panel member (the speaker struggled with the name; we label them Speaker X) emphasized:

  • Safety and diffusion must run in parallel – they are not mutually exclusive.
  • International cooperation is essential for high‑risk AI deployments (e.g., autonomous weapons, loss‑of‑control scenarios).

Speaker X referenced Professor Benjo’s research on “loss of control” risks and called for public‑institution‑led coordination rather than relying solely on the private sector.

6. The Code of Practice as a Reference Point

The moderator (or closing speaker) summarized the session’s core conclusion:

  • Innovation and trust can coexist if appropriate frameworks are institutionalised.
  • The code of practice – especially its safety chapters – should serve as a standard for other countries to adopt.

The audience was urged to review the code of practice and consider signing up to its guidelines.

7. Closing Call to Action

The final remarks reiterated:

  • Do not lose more time – leaders must sit down, use summit occasions, and make concrete progress.
  • Continue the discussion beyond the summit, with all stakeholders (government, industry, academia, civil society) collaborating on safe AI deployment.

The panel thanked participants, emphasized that the session was “very interesting,” and expressed optimism about future joint efforts.

Announcements

  • Presentation of the draft “Code of Practice” – highlighted during the session, with an invitation to examine the safety chapters.
  • No product launches or new initiatives were announced beyond the code of practice reference.

Key Insights

InsightSpeaker/Source
A co‑legislative “code of practice” is more adaptable than rigid legislation.Professor Benjo (referenced)
Governments must have enforcement capacity comparable to large private AI firms.Brando (moderator)
Global cooperation (EU, US, China) is essential for safety standards.Sean
Trust must be built around domain‑specific use‑cases rather than a one‑size‑fits‑all approach.Paula
Safety and diffusion should be pursued in parallel, not sequentially.Speaker X
The code of practice can become a benchmark for other nations.Closing speaker

Data & Findings

  • The transcript did not contain quantitative data, statistics, or empirical research results.
  • References were made to “existing compliance by many companies” with the code of practice, but no specific figures were provided.

Recommendations / Calls to Action

  1. Adopt and adapt the Code of Practice – especially the safety provisions – at national and regional levels.
  2. Create mechanisms for firms to safely reduce speed (e.g., “slow‑down” protocols) when approaching critical risk thresholds.
  3. Establish international forums where AI developers from all major regions can negotiate equal terms.
  4. Develop sector‑specific trust frameworks that define clear validation, testing, and monitoring standards.
  5. Public institutions should lead on high‑risk AI domains (military, autonomous systems) rather than relying on market forces alone.

Open Questions / Debates

QuestionStatus
How can a code of practice stay sufficiently flexible while remaining enforceable?Ongoing – needs continual stakeholder engagement.
What concrete slow‑down mechanisms can be implemented without harming competitiveness?Proposed but not yet defined.
How to ensure equitable participation of countries with differing AI capabilities (e.g., China vs. EU) in a global governance framework?Recognised as essential; details pending.
What are the precise trust controls for high‑stakes domains like healthcare?Needs sector‑specific standards development.
How to address loss‑of‑control and military AI risks under existing legal regimes?Identified as a gap; requires dedicated public‑institution action.

Key Takeaways

  • A “code of practice”—co‑legislated by governments, industry, academia and civil society—offers a pragmatic path to governing AI safely while preserving flexibility.
  • Governmental enforcement capacity must match that of large AI firms to ensure rules are applied uniformly and trust is built.
  • Global cooperation is non‑negotiable; all major AI‑producing regions must engage on an equal footing to set shared safety standards.
  • Context matters: Trust frameworks need to be tailored to specific use‑cases (e.g., medical AI vs. customer‑service chatbots).
  • Safety and diffusion are parallel goals; they can be pursued simultaneously through coordinated policy and industry action.
  • The draft code of practice’s safety chapters should be adopted as an international benchmark, inviting nations to sign up and harmonise standards.
  • Immediate leadership action is required—deliberate, time‑sensitive meetings should be convened to cement progress on safety, loss‑of‑control, and military AI concerns.
  • Ongoing dialogue among policymakers, scholars, and industry leaders is essential to refine and operationalise trustworthy AI at scale.

See Also:

Graph View

Backlinks