From reactive tools to digital colleagues: the rise of agentic AI

We are entering a new era of artificial intelligence that extends far beyond current generative AI applications. While today's AI primarily serves to write content, produce code, create visuals, or generate ideas, the next evolutionary step is agentic AI—systems capable of independently planning and executing sophisticated processes. This profound shift introduces new legal challenges as existing frameworks strain to address AI's increasing autonomy and decision-making capabilities.

Understanding Agentic AI

Agentic AI refers to AI systems endowed with agency—the ability to autonomously perceive, reason, act, and learn in pursuit of goals with minimal human oversight. These systems combine generative capabilities with decision logic and tool use, enabling them to "focus on doing" rather than just generating content.

Key characteristics include autonomous decision-making and proactiveness, where systems can assess situations and determine next steps with little to no human input. Once an agentic system understands a goal, such as "organise my calendar and book travel for a meeting," it can carry out necessary sub-tasks independently. Additionally, these systems employ goal-driven problem solving through a sense-plan-act cycle: observing data, reasoning about it, executing actions, and learning from results.

Traditional generative AI operates in a single-step input/output manner without persisting goals, focusing on creating content reactively in response to prompts. Agentic AI, by contrast, maintains a goal-oriented feedback loop and handles entire processes rather than single outputs. For example, while generative AI might write an email when asked, agentic AI could draft the email, send it, select a location, schedule meetings based on replies, and update calendars—all semi-autonomously.

Autonomous Software Development

Agentic AI is moving software development from automated code generation toward fully autonomous management of entire development cycles. These systems proactively analyse requirements, independently plan software architecture, execute comprehensive coding tasks, and autonomously conduct testing, debugging, and deployment with minimal human supervision. Such systems continuously interact with development environments, APIs, and external resources, dynamically adapting their approaches based on real-time feedback and changing project parameters. This autonomy significantly accelerates software delivery cycles and improves resource allocation.

Copyright Implications

The increasing autonomy of agentic AI intensifies the legal challenges to traditional copyright principles when compared with the use of “mere” generative AI.Current copyright law requires that works constitute "intellectual creation reflecting the freedom of choice and personality of its author" according to the European Court of Justice. This threshold becomes problematic when generative AI — and even more so agentic AI — autonomously makes the most defining creative decisions.

Unlike the use of generative AI for assistive tasks such as simple corrections, agentic AI operates as an autonomous agent, executing complex, multi-stage tasks with minimal human input. This shifts the human role from direct creative engagement to the mere definition of high-level goals.. When agentic AI independently produces entire software modules based solely on general requirements without meaningful human input, the level of human creativity needed for copyright protection may be no longer met.

For tech-focused companies that license code to third parties, the use of agentic AI raises fundamental strategic and legal questions. If AI-generated code cannot be protected by copyright, traditional licensing models may collapse—especially if clients can use similar AI tools to recreate the code themselves. To preserve value, companies must either ensure key software components remain authored by humans or rely on alternative protections such as trade secrets and confidentiality. If, instead, AI becomes the main driver of output, the business model may need to shift—away from licensing protected IP toward offering superior implementation know-how, scalable infrastructure, or domain-specific AI solutions. This shift also brings liability, regulatory compliance, and unfair competition law into sharper focus...

AI Act Compliance

Classification and regulatory implications under the AI Act become increasingly complex for agentic AI solutions. Companies currently face foundational questions about how individual AI tools should be classified (e.g., as General-Purpose AI models). For genuinely agentic systems, these challenges intensify significantly as they often operate through multilayered expert systems collaboratively tackling development tasks.

Real-Time Entertainment Content

Agentic AI is reshaping digital entertainment development, moving beyond traditional generative tools to systems that autonomously design, run, manage, and dynamically evolve content. Applications include acting as autonomous DJs generating real-time music, creating personalised TV series, or producing and managing video game content.

In the games industry, while reservations exist about delegating core IP creation to AI, developers are experimenting with agentic AI for generating entire casual games and using it selectively for established IP. Applications include managing monetisation processes, adapting gameplay in real-time to player behaviour, and generating content-rich elements like NPCs with real-time dialogue or dynamic environments.

Copyright Challenges in Gaming

Real-time content generation creates dual challenges: establishing protectability of AI-generated material and ensuring autonomous agents don't infringe third-party rights. Fully autonomous machine-generated content typically may not meet copyright protection thresholds. A future approach could involve relying on pre-cleared “design corpus” assets, whose existing copyrights may extend to AI-generated outputs.Furthermore, theprinciple may emerge that more important assets require tighter AI constraints, potentially even deterministic behaviour. For example, main characters might be reproduced exactly while AI has more autonomy in shaping environments.

When deploying agentic AI to generate real-time game content, developers face dual risks: securing rights in newly created outputs and avoiding copyright infringement from third-party material the AI may incorporate. If the agentic AI reproduces such material or communicates it to the public without authorization, this may pose risks under EU copyright law. This underscores the need for robust safeguards, especially since it is the agentic AI itself that autonomously creates the content. These safeguards include predefined content boundaries and human oversight for critical assets like characters or plot developments.

Where players can influence the outputs of agentic AI, a shift toward intermediary liability—based on the notice-and-action mechanism—may offer a more appropriate legal framework. For example, if players use the AI to generate infringing content (such as a copyrighted character from another game), and the game provider removes the content after being notified, liability could potentially be avoided—similar to how user-generated content is treated under existing rules.

Youth Protection and Media Regulations

The use of agentic AI to generate real-time content in games raises serious risks under youth protection laws. Developers must prevent the AI from producing prohibited content—such as child sexual abuse material, hate speech, or Nazi glorification—which could trigger direct liability and fines (up to EUR 500,000 in Germany). For age-restricted content, outputs must align with the game’s rating (e.g. USK or PEGI). This requires either global safeguards based on the strictest rating or region-specific age-gating—both of which demand technical constraints to block inappropriate content and avoid regulatory sanctions..

Real-time AI generation also challenges traditional age rating systems, which rely on full content review before release. Since agentic AI creates key elements dynamically, some content may not be assessable in advance—potentially invalidating the game’s rating. In jurisdictions where interaction risks affects classification, AI-generated content could even trigger higher age ratings. To stay compliant, developers may need to restrict the AI’s creative freedom by using pre-approved modules, vocabulary, and assets that can be reviewed and rated beforehand..

Future Outlook

In the short term (one to two years), experts anticipate rapid advancements in agentic AI with realistic acknowledgment of limitations. Expected developments include continued evolution from current prototypes to production-grade systems and broader enterprise adoption as companies scale from isolated pilots to wider deployment.

The rise of agentic AI represents a fundamental shift requiring new legal frameworks, business models, and regulatory approaches as AI systems transition from tools to autonomous digital colleagues capable of independent decision-making and execution across diverse sectors.