At first, I was afraid of the complications artificial intelligence (AI) could bring to the practice of transfer pricing (TP). I kept thinking about how TP professionals could adapt with the growing influence of AI.
AI is no longer a far-fetched phenomenon; it has transformed and continues to transform how businesses operate and compete. As it becomes more integrated into business models, important questions arise regarding intellectual property (IP) ownership and the pricing of intercompany transactions involving AI. These considerations affect how multinational enterprises (MNEs) allocate profits and mitigate TP risks. For MNEs, understanding the interplay of AI and TP is crucial for staying competitive and compliant.
AI systems contribute to value creation through proprietary algorithms, predictive analytics, or generative outputs. MNEs must evaluate how this value is shared and compensated. For instance, if a Philippine subsidiary uses an AI model developed by its parent company, does this give rise to royalties? If the subsidiary contributes data as part of developing or training the model, should there be compensation? These questions are central to TP and require careful analysis of functions, assets, and risks.
The Organisation for Economic Co-operation and Development (OECD) and Philippine TP Guidelines emphasize the arm’s-length principle, which assumes that independent parties price transactions to reflect the true economic value of each party’s contributions. When AI is involved, this becomes more complex, as the value chain may include intangible contributions that are difficult to quantify.
IP OWNERSHIP AND VALUATION
AI systems often involve complex IP structures, such as algorithms, training data, and model outputs. Determining who owns the IP and how it should be valued is critical. Guidance on intangibles under the Philippine TP rules and the OECD Guidelines may provide a framework, but AI introduces new challenges, especially when outputs are co-created by different entities or continuously updated.
Ownership of AI-generated IP may not be straightforward. For instance, if a Philippine entity contributes market data or domain expertise to train a global AI model, it may have a claim to a share of the resulting value. This raises questions about how to allocate returns and whether a cost contribution or licensing arrangement is appropriate.
COST CONTRIBUTION ARRANGEMENT (CCA)
When multiple entities contribute to AI development, whether through data, engineering, or domain expertise, a CCA may be appropriate. Recognized in the Philippine TP Audit Guidelines, CCAs allow group members to share the costs and risks of developing intangibles in proportion to their expected benefits.
However, implementing a CCA requires robust documentation, clear delineation of contributions, and reliable valuation methods. Philippine companies participating in such arrangements must ensure that these elements are properly addressed to reflect economic reality and comply with TP requirements.
CHARACTERIZATION OF AI-RELATED TRANSACTIONS
AI can blur the lines between services (e.g., customer service automation) and IP (e.g., proprietary algorithms, trained models). For example, AI may be embedded in broader service offerings, such as personalized recommendations in e-commerce. In such cases, TP analyses must consider whether the AI component should be treated as a distinct intangible and unbundled from the overall service, in order to determine the correct pricing and tax treatment.
It’s important to ensure that the contracts and corresponding TP documentation accurately reflect the nature of the transaction. Otherwise, if the language does not match the substance of the transaction, there is a risk of misclassification and unintended tax consequences. For example, there may be instances where intercompany agreements reference the use of IP, even if in reality, no IP is used. Unless further documentation shows otherwise, this may potentially be viewed as giving rise to royalties and taxed as such, despite the substance of the transaction being more aligned with a service. Clearly, reviewing the contractual provisions is crucial to managing TP risks.
BENCHMARKING
Finding comparables for AI-related transactions at this time can be difficult as they are still fairly new. Traditional benchmarking methods may not capture the full value of AI-driven activities, especially when intangibles are involved. These may include joint contributions from multiple entities, making it challenging to identify what value was created and by whom.
In such cases, alternative approaches like the profit split method may be more appropriate. Since it allocates profits based on each party’s role in the value creation process, it may be better suited for transactions involving unique intangibles and collaborative efforts that lack reliable market benchmarks.
BALANCING AI WITH PROFESSIONAL JUDGMENT
While posing challenges for TP, AI also offers benefits to the TP practice itself by improving efficiency in compliance processes, from automation and documentation to data analysis and risk assessments. That said, at this stage, AI tools would still need to be developed further. And even as it makes giant leaps, AI cannot fully replace professional judgment. Understanding the economic substance of transactions, interpreting regulatory guidance, and managing audit risks still require human skills and insight.
In particular, human expertise remains indispensable in interacting with tax authorities. TP professionals must be able to explain complex business models, defend positions during audits, and negotiate outcomes with both technical knowledge and interpersonal skills. These interactions often involve nuance and trust-building which are areas where AI may fall short.
The most effective approaches will be those that are human-led and tech-powered. This means leveraging AI tools to handle data-heavy tasks while relying on experienced professionals to provide oversight, contextual interpretation, and ethical decision-making.
Although the Philippines does not yet have AI-specific TP regulations, companies can already consider proactively addressing the TP implications of AI adoption. This includes updating TP documentation to incorporate AI-related functions, assets, and risks; reviewing intercompany agreements involving AI systems and IP to ensure they reflect economic substance; engaging cross-functional teams, including tax, legal, finance, and technology, to align on value creation and compliance strategies; and monitoring global developments, such as OECD guidance, to anticipate future regulatory shifts.
By tackling the TP challenges posed by AI, businesses can reduce risk, improve compliance, and position themselves for long-term success in a digital economy. Tax authorities may also consider these developments in legislation and policymaking as current laws and regulations may not fully account for AI-related transactions.
As technology advances rapidly, blending innovation with sound tax governance will be a key differentiator.
The views or opinions expressed in this article are solely those of the author and do not necessarily represent those of Isla Lipana & Co. The content is for general information purposes only, and should not be used as a substitute for specific advice.
Patrick Andrew Lim is an assistant manager at the Tax Services department of Isla Lipana & Co., the Philippine member firm of the PwC network.
+63 (2) 8845-2728
