Since January 1, 2026, the revised Patent Examination Guidelines have introduced specific examination standards and examples for AI- and big-data-related inventions under Article 5(1) of China’s Patent Law. Even where an AI-related application qualifies as a technical solution, it may still be excluded from patent protection if its data collection or decision-making logic raises legal or ethical concerns.

For autonomous driving technologies, two areas deserve particular attention: the collection of biometric data and decision-making logic in emergency situations. For companies developing and patenting autonomous driving technologies, addressing these risks requires more than simply understanding the relevant provisions. Potential issues should be considered at the R&D, patent drafting and examination-response stages.

1. Biometric Data: Does the Invention Comply with the Law?

Autonomous driving systems may routinely involve the collection or processing of biometric information. Typical examples include driver monitoring systems that analyze facial features to detect fatigue, external sensors that capture images of pedestrians or other drivers, and facial recognition systems used for vehicle unlocking or identity verification.

These applications may implicate not only patent law but also China’s personal information protection and automobile data regulations. In particular, inventions involving facial features, voiceprints, gait or other biometric characteristics may face increased risk where individuals are unaware of, or have not consented to, the collection or processing of such information. By contrast, scenarios falling within statutory exceptions for security or public safety may present relatively lower risks, while applications involving marketing or user profiling require greater caution.

Where necessary, the technical solution itself may be adjusted to reduce such risks. For example:

• Build authorization into the technical solution. Instead of describing data collection as an unqualified function, the system can include a user authorization or confirmation step, reflected in the specification and claims.
• Use anonymization or data minimization. Where precise identification is unnecessary, the system may extract relevant features and immediately delete the original biometric data.
• Narrow the application scenario. Where the invention is genuinely related to security or public safety, the relevant statutory purpose and technical implementation can be clearly reflected in the patent application.

The key point is that compliance should be incorporated into the technical architecture rather than treated merely as background legal language.

2. Emergency Decision-Making: Does the Logic Violate Public Morality?

The second, and potentially more difficult, issue concerns how an autonomous driving system allocates safety resources during emergencies.

The revised examination framework specifically addresses scenarios in which an emergency decision-making model distinguishes between people based on characteristics such as gender or age when determining whom to protect or whom to collide with. Such decision-making may be considered contrary to public morality. More broadly, the use of sensitive characteristics such as race or disability status as the basis for differentiated safety decisions may also present significant risks.

There are also potential analogous scenarios that have not yet been directly confirmed by official examination examples—for instance, prioritizing vehicles based on passenger numbers or vehicle value, assigning different right-of-way priorities according to vehicle type, or providing different levels of safety redundancy based on a user's membership status. These situations involve similar questions about whether identity, status or economic value should influence the allocation of safety resources. Their legal status is less certain than the official examples, however, and they should therefore be treated as potential risks rather than established prohibited categories.

Possible technical adjustments include:

• Replace sensitive attributes with neutral physical or behavioral parameters, such as body size, movement speed, posture stability or trajectory-prediction confidence.
• Change the optimization objective from protecting particular categories of people to minimizing overall collision energy or injury probability.
• Remove differentiated treatment altogether and retain only general risk-avoidance and path-planning strategies.

This type of adjustment can be more challenging because the differentiation logic may itself constitute an important part of the invention’s technical contribution.

3. Article 5(1) Cannot Be Considered in Isolation

Modifying a technical solution to address ethical concerns does not end the patent examination analysis. Article 5(1), inventive step and sufficiency of disclosure must be assessed separately and together.

Article 5(1) requires substantive adjustments to the relevant process or technical features. Inventive step requires an independent technical contribution, while sufficiency of disclosure requires appropriate technical details to enable implementation.

This interaction is particularly important when sensitive attributes in an emergency decision-making system are replaced. Once the technical features change, the scope of relevant prior art may also change. Applicants should therefore consider conducting a new prior-art search after modifying the solution, rather than assuming that merely replacing certain wording will resolve all examination issues.

4. Approaches to Office Action Response

Not every Article 5(1) issue requires a fundamental redesign. Depending on where the problem appears in the application, several alternative approaches may be available.

First, identify precisely where the risk arises. If the problematic ethical scenario appears only in a specific embodiment of the specification, while the claims are drafted at a sufficiently broad level, deleting that embodiment may provide a relatively straightforward solution.

Second, consider deleting a problematic dependent claim. Where the independent claim itself is unobjectionable, deleting the problematic dependent claim may be more efficient than pursuing a divisional application. A divisional application does not make a sensitive technical feature compliant by itself; its principal value may instead lie in separating the examination timelines of different claim sets.

Third, assess whether the problematic feature is technically necessary. If removing the sensitive attribute does not affect the technical problem being solved or the technical effect achieved, the applicant may argue that the feature is not an essential technical feature and seek its deletion.

Finally, consider alternative forms of protection. Where the core inventive value depends fundamentally on sensitive decision-making logic and there is insufficient room for amendment, protecting the relevant know-how as a trade secret or redesigning the technology and filing a new application may be more practical. Such decisions need not wait until an examination objection is received; they can be considered during the R&D stage.

5. Practical Takeaways for Autonomous Driving Companies

For autonomous driving companies, ethical risks under Article 5(1) should not be viewed as a one-time patent prosecution issue. They are structural considerations that can arise throughout the design of data collection, sensing and emergency decision-making systems.

A more effective strategy is to identify potentially problematic features at the R&D and patent drafting stages and incorporate appropriate compliance mechanisms or neutral decision-making criteria into the technical solution from the outset.

For applications already under examination, applicants should first determine exactly where the risk lies—whether in the claims, a specific embodiment or a particular technical feature—and then evaluate the most appropriate response, including amendment, deletion, argumentation, divisional filing, trade secret protection or a new application.

For autonomous driving technologies, patent strategy and compliance strategy are increasingly interconnected. Addressing ethical boundaries early can help companies reduce prosecution risks while preserving the commercial and technical value of their innovations.