Skip to main content
Perspective:
Perspective
25 Jun 2026
6 minute read

When AI acts on its own, who is responsible for mistakes?

Issue Averted: From AI autonomy to accountable action

By embedding legal oversight and clear accountability into agentic AI from the outset, businesses can harness autonomous technology with confidence and control.

Download the full report: Issue Averted - From AI autonomy to accountable action
Download PDF (917KB) View PDF

Build trust and accelerate innovation with agentic AI

Agentic AI is transforming business operations, enabling systems to act autonomously and deliver greater speed and efficiency. Yet, without clear accountability and robust governance, these advances can introduce legal, regulatory, and reputational risks. This Deloitte Legal report examines how early legal involvement, practical oversight, and cross-functional collaboration can help organizations deploy agentic AI responsibly—ensuring decisions remain explainable, outcomes are owned, and trust is maintained.

Embedding legal expertise from the outset, defining escalation paths, and establishing ongoing monitoring are essential steps. By mapping high-risk use cases, setting clear approval thresholds, and fostering teamwork between legal, technology, and business leaders, organizations can scale agentic AI with confidence and control.

Read the full report to discover how your legal team can turn agentic AI from a source of uncertainty into a foundation for sustainable innovation and competitive advantage.

  • Unclear accountability. Establish clear roles, review processes, and intervention points before deploying agentic AI. Legal teams should define who is responsible for decisions, set approval thresholds, and ensure human oversight is embedded throughout. This proactive approach prevents confusion and enables swift action when issues arise.
  • Regulatory and contractual exposure. Integrate legal and compliance expertise into the design and operation of agentic AI, especially in high-risk areas. Human-in-the-loop controls help avoid regulatory breaches, unfavorable contracts, and compliance failures. Early involvement and ongoing review ensure sensitive decisions are managed and risks contained.
  • Loss of trust. Build transparent governance and explainable AI outcomes to reassure customers, regulators, and partners. Show how decisions are made, what controls are in place, and where human judgement remains essential. Prioritizing transparency and accountability helps protect reputation and foster lasting trust in autonomous systems.

Download the report

  • Define accountability before deployment. Involve legal teams early to clarify what agentic AI systems are permitted to do, where human approval is required, and who owns each process. Establish escalation protocols and approval thresholds before launch, so responsibility is clear and governance gaps are avoided.
  • Embed legal into governance and design. Make legal a core part of the project team from the outset, not just a final checkpoint. Collaborate across legal, technology, and business functions to shape policies, record actions, and set controls for customer-facing or contract-related activities. Early legal input enables innovation while ensuring compliance.
  • Monitor agentic AI as an ongoing obligation. Treat oversight as a continuous process, not a one-off review. Build structured monitoring, regular reassessment, and adaptable policies into the operating model. As agentic AI evolves, legal teams should help maintain explainability, update governance, and respond to new risks and use cases.

Download the report

  • Reduced redesign and rework. Early legal involvement helps organizations adopt agentic AI faster by reducing late-stage redesigns, confusion, and avoidable risk. Surfacing risks earlier in the lifecycle reduces costly redesign and rework.
  • Improved trust and transparency. A well-designed human-in-the-loop approach helps combine speed with oversight, giving teams clearer control over where judgment, review, and intervention remain essential. Making autonomous decision-making easier to explain improves trust with stakeholders inside and outside the organization.
  • Sustainable scaling of AI. Legal builds governance that is practical, proportionate, and suited to systems acting with increasing independence. This enables sustainable scaling of AI across functions, so the business isn’t starting from scratch each time. Legal helps the business move forward with confidence.

Download the report

Intangibles, Data & Technology

As the commercial and legal significance of data, software, ideas, and intellectual property continues to accelerate, we will help you better understand, protect, develop, and capitalize on, these intangible assets.

Learn more

Did you find this useful?

Yes
No

Thanks for your feedback

Your feedback is important to us

To tell us what you think, please update your settings to accept analytics and performance cookies.

Need help updating cookies?

Get in touch

Melinda Upton

Melinda Upton

Partner, Tax & Legal
+44 7425 747904
Richard Punt

Richard Punt

Deloitte Global Leader | Legal Services
+44 20 7007 0509
Sebastiaan ter Wee

Sebastiaan ter Wee

Partner
+31630432661

Our thinking

Issue Averted: Finding the way forward when hackers get in

Transform cybersecurity crises into resilience and innovation by navigating chaos, mitigating damage, emerging stronger.
Research
5 minute read

Issue Averted: Flip supply chain readiness from burden to value

How to avoid supply chain breakdowns, reputational damage, and material fines as your organization faces expanding regulatory requirements and unpredictable global events
Research
5 minute read

The legal implications of Generative AI

The current enthusiasm for AI adoption is being fueled in part by the advent of Generative AI
Service
5 minute read

From vision to value: Scaling Generative AI deployment with speed and confidence

Generative AI presents a significant opportunity for businesses, but scaling responsibly requires navigating legal, ethical, and regulatory complexities.
Research
10 minute read