The rise of the machines: agentic AI The rise of the machines: agentic AI

If 2024 and 2025 were defined by the rise of large language models (LLMs) and the innovation they unlocked, 2026 is emerging as the year of agentic AI. In our 2024 Osler Legal Outlook article, Unlocking AI innovation, we explored how the first wave of generative AI showed how models could summarize, draft and analyze information. The next wave of generative AI will go significantly further and is expected to include systems built on top of LLMs that can plan and act autonomously. AI is evolving from simply a tool that can assist users, to an agent that can pursue goals and complete tasks on its own. 

This evolution also builds on broader developments in how AI is reshaping law and industry, from emerging copyright considerations with model training and deployment, to integration of AI in the health sector and AI’s emerging role in litigation.

The legal implications for intellectual property, data privacy, tax and commercial contracting are already emerging and will need to be a focus for legal and compliance teams moving into 2026 and beyond.

What is agentic AI?

Agentic AI refers to systems, typically software, that can plan, reason and take action towards a defined objective. Unlike traditional LLMs that must be prompted by a user, an agent is able to independently pursue an outcome. This process can involve multiple steps, including retrieving information, making decisions and taking necessary action. Agentic AI systems are much closer to a digital employee.

The potential use cases are extensive. In a call centre, an agent might — much like a human call centre agent — handle customer messages across email and chat, escalate complex issues and schedule follow-ups. In gaming, an agent may compete against human players by reasoning about how to achieve its objectives.

This shift from the use of LLMs to more goal-oriented, autonomous agents has prompted industry leaders to look beyond traditional software-as-a-service toward truly goal-driven systems. Over the past year, Google introduced its Agent Development Kit, an agent creation framework, and Cohere launched North, an agentic platform for businesses. These are just two examples of the agentic AI platforms that enterprises are beginning to adopt as they explore how to embed autonomous frameworks into their operations.

Legal implications

When software can plan, decide and act on its own, established assumptions about ownership, control and responsibility are challenged and require an organization’s legal and compliance approach to be accordingly adaptive.

Intellectual property

Agentic AI shifts the focus of intellectual property from models to the systems that make them act. The protectable value lies not only in model weights or prompts but also in the software orchestration layer, workflows, system prompts, connectors and adaptors. These evolving agentic intellectual property assets will represent an organization’s digital crown jewels.

Protecting and commercializing these assets will require a layered approach across copyright, patent, trade-secret and even commercial contracting. Contracts should be explicit about ownership and other rights relating to these assets. Appropriate limitations on use, modification and access by others, as well as licensed rights generally, need to be properly scoped.

These intellectual property elements inform every other legal issue that flows from the development and use of agentic AI — from privacy, to contracting, and even to tax. Organizations must consider which parts of their agentic stack they own, how to secure them and how to properly address and allocate any associated risks to the business.

Privacy and data protection

Agentic AI amplifies existing privacy and data protection risks associated with AI technologies in new ways. As agents autonomously gather, transmit and analyze data across systems, additional complexities may arise that organizations, developers and users must consider. These can include the possibility that data and personal information processed by agents will be subject to multiple privacy and data protection laws and that this information may be used in ways that are unanticipated, including to make automated decisions.

Among other things, organizations must consider and understand the data flows associated with autonomous agents, including how data moves between systems and across borders. Unlike traditional systems, agents acting autonomously may exchange data with other agents and adapt their behaviour over time. Organizations must also ensure that individuals whose information may be processed by an agent understand when and how agents may collect and use personal information, particularly as autonomous and multi-agent interactions make consent and transparency more difficult to manage.

It will also be important for organizations to implement appropriate privacy and security contractual protections when engaging third parties that may use agents to process data or personal information on behalf of the organization.

Autonomous delegation will not dilute accountability. The deploying organization remains responsible for lawful, fair and transparent processing, even when decisions are made by autonomous systems.

Tax considerations

As agents begin performing work across jurisdictions, tax counsel will need to address and apply long-standing principles against this new backdrop. If an AI agent is operating from a device in Canada, an organization must assess whether this could create a permanent establishment or other taxable nexus in Canada for a foreign enterprise. Relatedly, income generated by AI-driven services may need to be characterized and allocated differently than previously reported as a result. Reductions in human headcount in any particular jurisdiction may also affect how income is taxed. While agentic AI may provide flexibility in workflows and allocations, it may also trigger the need to assess the organization’s transfer pricing model.

These questions become more relevant as agentic systems move from pilot programs to active deployment. Organizations should monitor legal developments and local tax authority and OECD guidance and take steps to ensure business models, corporate structures, and intercompany and commercial arrangements can withstand scrutiny. As agentic AI changes how businesses create and deliver value, tax counsel may need to rethink the related implications.

Commercial contracting

The rise of agentic AI is prompting organizations to revisit commercial contracting. Some organizations are evaluating whether current procurement and contractual frameworks are suitable, as traditional service contracts assume a human service provider. By contrast, agentic arrangements may have to allocate ownership and risk to reflect that actions will be taken by automatous software. These same contracts will need to clearly define ownership and use of outputs and underlying data flows for multiple components of the agentic solution. They will need to account for constant data and software generation and sharing that introduces new intellectual property elements.

The use of agentic AI may also require reconsideration of contractual risk allocation, including indemnification and limitations of liability, to address agent actions that are both within and outside of its intended scope. The focus should be on allocating risk to the party best able to control the risk. This allocation may differ from risk allocations for traditional software applications.  

Contracts will need to evolve from managing tools to managing human-like actors. It is not yet entirely clear how this will develop.

How widespread will agentic adoption be in 2026?

The coming year will likely mark a turning point for how organizations integrate agent autonomy into their operations. Just as the past two years were spent understanding and regulating LLMs, the next phase will focus on deploying, governing and contracting for AI agents.

Approaches to traditional legal frameworks will need to adapt. For legal teams, this means working closely with technical and business leaders to modify agreements, approaches to intellectual property, privacy, contracting and tax, as well as related compliance processes to address technologies that learn, act and evolve. These questions are no longer speculative; they are emerging in procurement and contract negotiations today.

We anticipate that 2026 will not simply mark the next phase of AI adoption, but scaling and growth of the agentic era. Organizations that prepare early will be best positioned to unlock the business transformation that we expect to unfold.