EPO Examiners Point Out to the Heads of Delegations in the Administrative Council of the EPO That the "AI Policy" of the Office is Illegal
There's an open letter circulating this month, reaching out not only to EPO staff but also to national delegates who enabled Benoît Battistelli and António Campinos. They reciprocated for bribery. That's how crooked this whole system of governance is. It's just blatant corruption; they even bribe the media and academia to cover up what's going on.
This time the subject is slop, or how the EPO management wants to replace staff with slop, as previously covered in:
- EPO Workers Caution That the Officials Are Still Illegally Trying to Replace Staff With Slop (to Lower Quality and Validity of European Patents)
- EPO Likely Breaking the Law Yet Again, This Time by Using Slop for Patents (to Lower Costs While Producing Monopolies That Cause Ruinous Lawsuits)
- EPO Poll: 68% Dissatisfied With Quality of Slop (Wrongly Framed as "AI") for Patent Classification
- The EPO, Europe's Largest Patent Office, Admits Outsourcing to Microsoft Slop
- European Patent Office (EPO) Illegally Turning to Slop Behind Closed Doors, Staff Objects to This Hidden Catastrophe
Not only does today's EPO grant many European software patents - i.e. patents which are both illegal and undesirable - it also wants to automate this process using software.
"In an open letter sent on 17 June 2025 to the Heads of Delegations in the Administrative Council," says the Central Staff Committee about its own letter, "the Central Staff Committee (CSC) asks the Administrative Council to exert its supervisory role and instruct EPO management to enter into genuine dialogue with the staff representation on the AI Policy, to revise the “Leverage AI” target of 90% AI-automated classification in the SP2028 and to put in place the measures supported by staff in the resolution."
They've put that in an "open letter with the adopted resolution as annex." (We replicated this annex here before, but we'll do so again)
Today we reproduce it below:
European Patent Office
80298 Munich
GermanyCentral Staff Committee
Comité central du personnel
Zentraler PersonalausschusscentralSTCOM@epo.org
Reference: sc25038cl
Date: 13/06/2025
European Patent Office | 80298 MUNICH | GERMANY
To: Heads of Delegations in the Administrative Council
Cc: Mr António Campinos (President of the Office)By email:
To: council@epo.org
Cc: president@epo.orgOPEN LETTER
Call for revising the EPO AI Policy and the “Leveraging AI” SP2028 objective
Dear Heads of Delegations in the Administrative Council, Staff of the European Patent Office gathered on 5 June 2025 in a General Assembly attended by 906 participants1 for Munich and The Hague, and 53 in Berlin.
A resolution on the EPO AI Policy2 was put to a vote and adopted by an overwhelming majority3: 638 voted in favour, 32 against and 23 abstained. The outcome confirms the concerns of staff on the EPO AI Policy.
The EPO announced its AI Policy on 5 March 2025 via an Intranet news item without consultation of staff and their representation, and without any decision from the Administrative Council. In the Technical and Operational Support Committee (TOSC) meeting on 29 and 30 April, the EPO made a presentation4 on the AI Policy but did not table the policy itself. The Member States delegations only received a link to the policy document on the Single Access Portal the day before the meeting.
________
1 906 participants connected into the meeting with a maximum of 736 simultaneous connections
2 “Resolution adopted on EPO AI Policy” [Annex]
3 585 (MU/TH)+53(BE) in favour, 32 (MU/TH)+0 (BE) against, 23 (MU/TH) abstained
4 CA/T 10/25
The EPO AI Policy is not bound by international texts and just takes note of initiatives such as the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy, and the Rule of Law of 17 May 2024, and the EU AI Act. While the EU-AI Act (Recital 6) defines “human-centric AI” as having the aim of increasing human well-being, the EPO restricts its “human-centric” approach as “final decisions at the EPO being taken by human actors”. This is insufficient. If there is no systematic human review of intermediate steps, AI influences the final decision and humans lose the control. As a consequence, the EPO AI Policy lacks boundary between “assistance” and “decision- making”.
A Key Performance Indicator (KPI) “Leverage AI” was mentioned for the first time in the third quarter of 2024 in CA/105/24 and is defined in the Strategic Plan 2028 Progress Dashboard. The KPI measures the proportion of classification tasks that no longer require human intellectual classification (thanks to AI support), compared to the total number of classification tasks. The target is 90% by 2028. Instead of a mere “optimisation of human intervention”, it is a (quasi) supplanting of humans which is taking place in classification and a loss of expertise built over decades. It is even highly debatable whether 10% of fresh human intellectual classification would be sufficient for AI model training in the future.
A poll run in February/March by the Central Staff Committee (CSC) among gérants (classification experts) gathered 247 responses and showed that 67.61% of them are dissatisfied with the quality of the AI classification. Accurate classification is essential for the quality of the search and examination. It is also a service to the users of the patent system and for the public.
Finally, the EPO explicitly does not exclude using AI systems in employment matters: for recruitment, promotions and termination of service; for evaluating the eligibility of individuals for essential benefits and services, including healthcare coverage; for assessing the likelihood that an individual will engage in misconduct and in the course of investigations. Such AI systems bear the risk of categorisation, profiling and algorithmic bias, and should not become a substitute to human management.
We can only ask the Administrative Council to exert its supervisory role and instruct EPO management to enter into genuine dialogue with the staff representation on the AI Policy, to revise the “Leverage AI” target of 90% AI-automated classification in the SP2028 and to put in place the measures supported by staff in the resolution.
Sincerely yours,
Derek Kelly
Chairman of the Central Staff Committee
RESOLUTION adopted on EPO AI Policy
The staff of the EPO, gathered in a General Assembly,
Noting that:
• The President of the EPO adopted a Policy on the use of AI systems in February 2025 without consulting staff nor their representation;
• The EPO hides behind its administrative autonomy under Article 4(1) EPC to consider international texts on AI (e.g. EU AI Act) as not legally binding;
• The EPO defines its “human-centric” approach as “final decisions at the EPO being taken by human actors” while the EU AI Act (Recital 6) defines human- centric AI as having the aim of increasing human well-being;
• The EPO explicitly does not exclude using AI systems in employment matters:
▪ for recruitment, promotions and termination of service1,
▪ for evaluating the eligibility of individuals for essential benefits and services, including healthcare coverage2,
▪ for assessing the likelihood that an individual will engage in misconduct and in the course of investigations3
• In a poll run among gérants, 247 responded and 67.61% of them replied they are dissatisfied with the quality of classification provided by AI-autoclose;
• AI systems hold the potential to support EPO staff in reducing routine workload and enhancing service quality — provided they are implemented transparently, responsibly, and with genuine consultation and training.
Fearing:
• the loss of human- oversight in intermediate steps and the resulting influence of AI in the final decision(s);
• the categorisation and profiling of EPO employees by AI systems;
• a detrimental effect on the quality of services provided by the EPO;
__________
1 page 12, EPO AI POLICY, Appendix A: Impact assessment and risk management, section B, 2c(ii)
2 page 12, EPO AI POLICY, Appendix A: Impact assessment and risk management, section B, 2d(i)
3 page 12, EPO AI POLICY, Appendix A: Impact assessment and risk management, section B, 2e
Requests the President and the Administrative Council to:
• bring the EPO AI Policy in line with the EU AI Act;
• revise the SP2028 goal aiming at target proportion of 90% of classification tasks that no longer require human intellectual classification;
• ensure that AI is not used as a substitute for human management;
• systematically consult staff and their representation prior to any implementation of AI;
• provide comprehensive training and dedicated time budget for staff to master AI tools relevant to their functions;
• transparently disclose where and how AI systems are implemented (model used, training and dataset used);
The Hague, Munich and Berlin, 5 June 2025
Important reminder: the recipients of this letter are being paid to play along with those who keep pressing with their own illegal proposals. It's a broken system, much like the notorious (annual) NDAA vote in the US.
As noted above, the media is paid not to talk about this stuff or to instead relay a bunch of puff pieces for the EPO. █