By Ana Herrera, Director of Patents, Technological Development and Innovation and Guillermo Criado, Director of International Business Development
Last December, the Japan Patent Office registered nearly 82,000 applications in a single month, well above the usual 20,000 or 30,000. In the absence of a convincing official explanation, a disturbing theory is gaining traction in the intellectual property sector: the use of artificial intelligence to generate applications automatically on a large scale.
The potential impact is clear. A thorough assessment of a patent application takes at least thirty minutes. Fifty thousand additional applications amount to 25,000 hours of work: more than twelve years of full-time work condensed into a few weeks. If this trend becomes established, it is not a statistical anomaly, but a warning sign of the pressure that the global patent system may come under.
At the same time, the Spanish Patent and Trademark Office (SPTO) has issued a notice that extends beyond our borders. It detects an increase in applications suspected of having been generated by AI and highlights two main risks.
The first risk is a security issue. When an inventor enters details of their invention into an artificial intelligence tool, the information may slip beyond their control. According to the terms of use, this data may be used to train models, stored on external servers, or even disclosed to third parties. In the field of patents, where novelty is an absolute requirement, premature disclosure can irreversibly compromise the protection of the invention or open the door to future challenges.
The second risk is a quality issue. An AI-generated application may appear to be formally correct, yet be legally ineffective. Inconsistent descriptions, vague definitions of technical elements or poorly formulated claims may lead to objections during examination or, worse still, to insufficient protection. A poorly drafted patent not only fails to provide protection, but also incurs costs, delays decision-making and gives a false sense of security. In many cases, it is more harmful than not having sought protection.
The European Patent Office (EPO) has also integrated artificial intelligence tools into its internal processes, for example in prior art searches and document classification. However, the office has been clear in its warnings. Its AI use policy, published in 2025, sets out a human-centred approach and emphasises the need to assess the legal, ethical and privacy risks associated with these technologies. Furthermore, the Guidelines for Examination in force since April 2026 state that ultimate responsibility for the content of an application always lies with the applicant and their representatives.
The underlying problem is not the occasional misuse of artificial intelligence by a company. It is the risk that the automated volume logic will distort the functioning of the system. If the mass filing of automatically generated applications becomes widespread, the system could become inundated with documents of little technical value. The result would be longer examination times, higher operating costs and a general loss of efficiency.
The consequences directly affect companies that are genuinely innovative. In a saturated market, those who invest years in research would be at a competitive disadvantage compared to those who churn out applications at a marginal cost. The parallel with inflation is clear: when something ceases to be scarce, its value diminishes. In this case, the value of the patent as a means of protecting and encouraging innovation.
It should be borne in mind that, in many sectors, a company’s patent portfolio is its main intangible asset. It underpins funding rounds, facilitates licensing agreements and defines the company’s competitive position vis-à-vis third parties. A portfolio built on automated documents without expert oversight can prove to be vulnerable when subjected to due diligence or litigation.
Drafting a patent is not merely a matter of technical writing, but a strategic decision. This involves defining what is protected and what is not, refining the scope of the claims, anticipating objections, and aligning the protection with business objectives. Artificial intelligence can be useful as a support tool, but it is no substitute for expert judgement. Without such oversight, an automated draft can become a source of risk.
A regulatory response is expected shortly, both in Japan and Europe and across the OECD as a whole. The question is whether it will do so before the impact becomes significant. In the meantime, companies should adopt a cautious approach and treat their intellectual property with the rigour it demands.
AI is already transforming the way work is done in the field of patents: it speeds up searches, improves freedom-to-operate analysis and optimises the management of global portfolios. When used under expert oversight, it delivers efficiency and value. However, when used as a substitute for legal and technical expertise, it can compromise the quality and the very protection of that innovation.
The patent system is based on a balance: society grants temporary exclusive rights in exchange for rigorous technical disclosure. Maintaining that balance in the age of AI will require upholding standards of quality, accountability and control.
