Responsible AI in Practice: Two Conversations, One Critical Question

At LegalTechTalk, two very different conversations converged on a single, urgent theme: Responsible AI is no longer theoretical—it is operational.

Across both a hands-on workshop and a policy-focused panel, one question kept resurfacing:How do we build AI systems that are not just powerful, but trustworthy?

From Theory to Practice: Building Boundaries Inside Legal Work

The first conversation unfolded in a workshop led by the Indian LegalTech Network (ILTN), co-facilitated by Shreya Vajpei, Siddharth Peter de Souza, and Nimrat Dhillon.

Titled “Building Your Own Boundaries: Responsible AI in Legal Practice,” the session moved beyond abstract principles and into the realities of day-to-day legal work.

The focus was simple—but difficult:

• How are legal professionals actually using AI today?

• Where do ethical dilemmas emerge in real workflows?

• And how can organisations design internal playbooks that balance efficiency with accountability?

  
  

What stood out was not just the diversity of perspectives but the honesty of the discussion. Participants didn’t just talk about possibilities—they unpacked risks, trade-offs, and the often invisible decisions that shape how AI is deployed in practice.

The takeaway was clear: Responsible AI cannot be outsourced to policy documents—it must be embedded into everyday decision-making.

---

From Practice to Policy: The EU AI Act and the Limits of Regulation

The second conversation shifted the lens from internal practice to external regulation.

In a panel discussion with Kirsty Harrower and Tom Whittaker, the focus turned to the EU AI Act—arguably the most ambitious attempt yet to regulate artificial intelligence at scale.

But instead of treating the Act as a checklist, the conversation interrogated its deeper implications:

• What does compliance actually look like in practice?

• Where do “dos and don’ts” fall short in complex legal workflows?

• And perhaps most importantly, should AI be governed this way at all?

  
  

The discussion made one tension unavoidable:

Regulation is necessary—but over-regulation risks stifling innovation.

The challenge is not simply to create guardrails, but to design ones that are flexible enough to evolve alongside the technology they aim to control.

---

The Common Thread: Responsibility as Design, Not Constraint

Across both sessions—one grounded in practice, the other in policy—a shared insight emerged:

Responsible AI is not about limiting what we can do. It is about defining how we choose to do it.

For legal professionals, this means:

• Moving beyond passive tool usage to active oversight

• Building internal frameworks, not just relying on external regulation

• Understanding that accountability cannot be automated—even if workflows are

  
  

The conversation is no longer about whether AI should be used in law. That question has already been answered.

The real question now is: How do we ensure it is used well?

---

Looking Ahead

LegalTechTalk made one thing clear—this is not a settled conversation.

Responsible AI will continue to evolve across jurisdictions, institutions, and workflows. And its success will depend not just on regulation or technology, but on the people designing, deploying, and questioning it.

With contributions from the ILTN community and support from the LegalTechTalk team—especially Bradley Collins and Merlin Beyts—these conversations are only just beginning.

Because in the end, building the future of legal tech is not just about innovation.

It is about responsibility.