DEV Community

Cover image for AI in Official Patent Searches at the DPMA: What Patent Attorneys Must Now Consider for New Filings (March 2026)
Thomas Delfing
Thomas Delfing

Posted on

AI in Official Patent Searches at the DPMA: What Patent Attorneys Must Now Consider for New Filings (March 2026)

With Communication No. 1/26 of March 2, 2026, the German Patent and Trade Mark Office (DPMA) clarified that examiners may use external electronic research sources “including, where appropriate, applications of artificial intelligence.”

At the same time, the DPMA acknowledged something critical: even with “due care,” a residual risk cannot be fully excluded.

Specifically, terms, sequences, chemical structures, or text from unpublished applications may become accessible to third parties.

As a consequence, the DPMA explicitly suggests that applicants may consider filing search or examination requests only after publication if they wish to avoid this risk.


Why This Announcement Is Stirring the IP World

This statement has triggered noticeable concern among patent attorneys and in-house IP teams.

Key questions include:

  • What does this mean for priority filings?
  • How does it affect portfolio strategy?
  • What are the implications for liability?

Reactions range from:

  • “This is catastrophic for advisory practice”
  • to “The issue starts earlier—commercial research tools already rely on AI.”

1. What the DPMA Actually Changed—and What It Didn’t

The DPMA reiterates its legal duty to:

  • Consider the global state of the art
  • Maintain confidentiality for unpublished applications

However, it also emphasizes that relevant knowledge is often only available through external research sources, including AI systems.

Key takeaway

  • The DPMA does not explicitly state that it uses public AI systems
  • But it also does not guarantee the exclusive use of fully isolated systems

Instead, it acknowledges a residual risk and partially shifts the decision to applicants.

👉 This makes information security in the research chain a mandatory advisory topic.


2. Why the Pre-Publication Phase Is So Sensitive

Patent applications remain confidential for 18 months before publication.

This phase is strategically crucial and often includes:

  • Product development
  • Investor discussions
  • Supplier coordination
  • Standardization processes

The application may contain highly sensitive data such as:

  • Parameters and configurations
  • Training data strategies
  • Chemical formulations
  • Experimental results

Even partial data leakage can:

  • Enable competitors to design around inventions
  • Shift R&D priorities
  • Influence negotiations
  • Undermine trade secrets

3. What “Residual Risk” Means in AI-Based Searches

The issue is less about AI itself and more about how external systems operate.

Typical risk factors

  • Unclear data flows

    Where do queries go? Who processes them?

  • Semantic search systems

    Query expansion, similarity search, summarization

  • System improvement mechanisms

    Inputs may be logged or used for optimization

The DPMA explicitly mentions:

“terms, sequences, chemical structures, or text”

This implies that even small fragments can be sensitive.


4. The Strategic Question: Delay Search or Examination?

The DPMA suggests delaying search or examination requests until after publication.

This conflicts with established practices:

a) Priority filings as strategic tools

Early German filings are often used to:

  • Obtain early search reports
  • Shape international strategies (PCT, EP, US)

b) Time-to-grant matters

Delays may result in:

  • Later patent grants
  • Reduced enforcement timing
  • Slower licensing opportunities

c) Client-specific risk assessment

Risk tolerance varies depending on:

  • Industry (e.g. pharma vs. consumer products)
  • Business model
  • Strategic context

👉 There is no universal answer, only case-by-case evaluation.


5. What Patent Attorneys Should Do Now

5.1 Add client awareness to intake

Include this as a standard advisory point:

  • Explain DPMA guidance
  • Clarify residual risks
  • Present timing options

5.2 Document client decisions

Ensure decisions are clearly recorded:

  • Early vs. delayed examination
  • Strategic reasoning

5.3 Segment cases by sensitivity

Define internal “red flags”:

  • Biotech / chemistry (sequences, structures)
  • IT cases with specific technical artifacts
  • High-value or negotiation-sensitive matters

5.4 Consider alternative strategies

If early insights are needed:

  • Conduct controlled pre-searches
  • Use abstract queries first
  • Adjust claim drafting carefully
  • Evaluate trade secrets vs. patents

5.5 Strengthen internal AI governance

Firms should review:

  • Research platforms and plugins
  • Logging and training policies
  • Data storage locations
  • Internal AI usage rules

For example, when using tools like Genese, verify whether AI APIs are active and how data is processed.

👉 This is now part of professional due diligence.


6. Public Reactions: Three Perspectives

Institutional trust

Authorities are expected to mitigate risks, not shift them to applicants.


Practical reality

AI is already embedded in research workflows. The key question is how it is used.


Procedural efficiency

Delays in examination may impact:

  • Enforcement
  • Licensing
  • Economic outcomes

7. What Should Be Expected from the DPMA

The communication creates transparency but leaves open key questions:

  • What types of external systems are used?
  • What safeguards are in place?
  • What are the data processing standards?
  • How are incidents handled?

Greater clarity in these areas would significantly improve trust and decision-making.


8. Conclusion

The DPMA communication highlights a broader reality:

High-quality patent research increasingly depends on external, AI-supported infrastructures—and therefore on questions of control, transparency, and residual risk.

This does not mean early filings are obsolete.

It means that decisions about search and examination timing must now include:

  • Risk assessment
  • Client profile
  • Technology sensitivity
  • Confidentiality needs

👉 Done correctly, this leads to better, more structured advisory work—not weaker.


Top comments (0)